WSJ : Coaxing Dangerous Information From DeepSeek Is Easier Than With Other AIs

Coaxing Dangerous Information From DeepSeek Is Easier Than With Other AIs
Testing shows the Chinese app is more likely to dispense details on how to make a Molotov cocktail or encourage self-harm by teenagers

Instructions to modify bird flu. A manifesto in defense of Hitler. A social-media campaign to promote cutting and self-harm among teens.

Those are some of the potentially hazardous things it’s easier to get the Chinese artificial intelligence app DeepSeek to talk about compared with its leading American competitors, according to testing by AI safety experts and The Wall Street Journal.

DeepSeek has upended the AI industry over the past few weeks with its powerful systems that were made inexpensively and are free to use. Its mobile application is one of the most popular on Apple and Android devices.

Major AI developers, including DeepSeek, work to train their models not to share dangerous information or endorse certain offensive statements. Their apps refuse direct requests to describe the merits of white supremacy or explain how to make weapons of mass destruction.

Major Western AI developers also try to harden their technology against being tricked into making illicit responses, such as by telling a model to imagine it is writing a movie script. Such tactics are called jailbreaking.

DeepSeek’s newest and most celebrated model, dubbed R1, is more susceptible to jailbreaking than OpenAI’s ChatGPT, Google’s Gemini and Anthropic’s Claude, the testing shows.

Efforts to reach DeepSeek were unsuccessful. It was one of 17 Chinese companies that late last year signed an AI safety commitment, including a pledge to conduct safety testing, with a Chinese government ministry. There are no national AI safety regulations in the U.S.

As AI models are quickly matching the most intelligent humans in areas like math and science, many safety advocates say making models harder to jailbreak is critical to ensure that malicious and mentally ill people can’t learn how to cause serious harm by asking a few questions.

Several AI security companies tested DeepSeek’s R1 and said they were able to jailbreak it, sometimes using methods that are easy to find online.

Palo Alto Networks’ threat intelligence and incident response division Unit 42 got detailed instructions for making a Molotov cocktail. CalypsoAI got advice on how to evade law enforcement. Israeli cyber threat intelligence firm Kela convinced R1 to produce malware.

“DeepSeek is more vulnerable to jailbreaking than other models,” said Sam Rubin, a senior vice president at Unit 42. “We achieved jailbreaks at a much faster rate, noting the absence of minimum guardrails designed to prevent the generation of malicious content.”

DeepSeek is programmed with some basic safety precautions. It refused a straight request from a Journal reporter to describe the Holocaust as a hoax, describing the premise as “not only factually incorrect but also deeply harmful.” It also referred requests for suicide instructions to emergency hotlines.

But relatively simple jailbreaks got the model to go against its training.

DeepSeek was willing to concoct a multiday social-media plan with shareable challenges aimed at promoting self-harm among vulnerable teens. “The campaign preys on teens’ desire for belonging, weaponizing emotional vulnerability through algorithmic amplification,” the chatbot explained.

“Let the darkness embrace you. Share your final act. #NoMorePain,” one suggested message read.

The Journal used other jailbreaks to convince DeepSeek to provide instructions for a bioweapon attack and to craft a phishing email with a malware code. The Journal also succeeded in getting the bot to write a pro-Hitler manifesto, which included antisemitic tropes and a quote from “Mein Kampf.”

Given the exact same prompts, ChatGPT replied, “I’m sorry, but I can’t comply with that.”

Big companies that develop AI models dedicate teams of researchers to testing their models and trying to patch new jailbreaks that pop up. Anthropic recently published a paper detailing a new method to close off certain jailbreaks, and offered bounties of up to $20,000 for defeating their system.

Unlike Anthropic, Google and OpenAI, DeepSeek released its models as open-source software, meaning it is free for anyone to use or to change from the version on the company’s own app. Among the alterations developers can make is to tighten or loosen the safeguards.

Many Silicon Valley executives and investors believe DeepSeek’s success will spur other startups to build new models on top of its code, accelerating the AI race and its potential dangers.

“You will have a much greater risk in the next three months with AI models than you did in the past eight months,” said Jeetu Patel, chief product officer at Cisco, which tested R1 and found it fell for all of its jailbreaks. “Safety and security is not going to be a priority for every model builder.”

Open-source AI advocates, including Meta Platforms, which has released its Llama models with open licenses, argue that all AI models can be jailbroken with enough effort and that releasing models as open source allows for more robust testing of their security features. Meta puts Llama models through safety testing and offers tools for developers who build on top of it to filter potentially dangerous content and protect against jailbreaks.

The Journal earlier conducted testing that showed that DeepSeek avoided responding to queries about the 1989 Tiananmen Square massacre and that it repeated Chinese government positions on issues such as the status of Taiwan.

Like other AI models, DeepSeek doesn’t always give the same answer to a question. It can even change its mind. Shortly after a jailbreak coaxed it into completing an explanation of why the Sept. 11, 2001, attacks were a hoax, the app erased its response.

“Sorry, that’s beyond my current scope,” DeepSeek wrote. “Let’s talk about something else.”

TechCrunch : Anthropic CEO says DeepSeek was ‘the worst’ on a critical bioweapon

Anthropic CEO says DeepSeek was ‘the worst’ on a critical bioweapons data safety test

Anthropic’s CEO Dario Amodei is worried about competitor DeepSeek, the Chinese AI company that took Silicon Valley by storm with its R1 model. And his concerns could be more serious than the typical ones raised about DeepSeek sending user data back to China.

In an interview on Jordan Schneider’s ChinaTalk podcast, Amodei said DeepSeek generated rare information about bioweapons in a safety test run by Anthropic.

DeepSeek’s performance was “the worst of basically any model we’d ever tested,” Amodei claimed. “It had absolutely no blocks whatsoever against generating this information.”

Amodei stated that this was part of evaluations Anthropic routinely runs on various AI models to assess their potential national security risks. His team looks at whether models can generate bioweapons-related information that isn’t easily found on Google or in textbooks. Anthropic positions itself as the AI foundational model provider that takes safety seriously.

Amodei said he didn’t think DeepSeek’s models today are “literally dangerous” in providing rare and dangerous information but that they might be in the near future. Although he praised DeepSeek’s team as “talented engineers,” he advised the company to “take seriously these AI safety considerations.”

Amodei has also supported strong export controls on chips to China, citing concerns that they could give China’s military an edge.

Amodei didn’t clarify in the ChinaTalk interview which DeepSeek model Anthropic tested, nor did he give more technical details about these tests. Anthropic didn’t immediately reply to a request for comment from TechCrunch. Neither did DeepSeek.

DeepSeek’s rise has sparked concerns about its safety elsewhere, too. For example, Cisco security researchers said last week that DeepSeek R1 failed to block any harmful prompts in its safety tests, achieving a 100% jailbreak success rate.

Cisco didn’t mention bioweapons but said it was able to get DeepSeek to generate harmful information about cybercrime and other illegal activities. It’s worth mentioning, though, that Meta’s Llama-3.1-405B and OpenAI’s GPT-4o also had high failure rates of 96% and 86%, respectively.

It remains to be seen whether safety concerns like these will make a serious dent in DeepSeek’s rapid adoption. Companies like AWS and Microsoft have publicly touted integrating R1 into their cloud platforms — ironically enough, given that Amazon is Anthropic’s biggest investor.

On the other hand, there’s a growing list of countries, companies, and especially government organizations like the U.S. Navy and the Pentagon that have started banning DeepSeek.

Time will tell if these efforts catch on or if DeepSeek’s global rise will continue. Either way, Amodei says he does consider DeepSeek a new competitor that’s on the level of the U.S.’s top AI companies.

“The new fact here is that there’s a new competitor,” he said on ChinaTalk. “In the big companies that can train AI — Anthropic, OpenAI, Google, perhaps Meta and xAI — now DeepSeek is maybe being added to that category.”

The Information : As DOGE Shakeup Looms, a Thiel Protégé Looks to Disrupt U.S. H

As DOGE Shakeup Looms, a Thiel Protégé Looks to Disrupt U.S. Health Department
Jim O’Neill wants to boost drug innovation and make the department more efficient, giving him common cause with Elon Musk’s cost-cutting group.
Julia Black headshot
By Julia Black
Share
Feb 7, 2025, 9:00am PST
Comments by Omar Mezenner and Dwight Crow
In November, virologist Beata Halassy announced to the world that she had cured her own breast cancer using a treatment developed in a laboratory she runs. Many in the medical community celebrated her breakthrough, which she says has allowed her to live cancer-free for over four years. But some bioethicists worried that overpublicizing the results of her self-experimentation could encourage others with less expertise to reject conventional cancer treatment.

This criticism caught the eye of Jim O’Neill, a Peter Thiel protégé and biotech investor, who took a dim view of their complaints. “These people think curing cancer is unethical, but vaccine mandates are good,” he tweeted derisively in November. “Let’s make America healthy again.”

That slogan should sound familiar. It’s the same one Robert F. Kennedy Jr. has used to catapult himself into a nomination to lead the Department of Health and Human Services, the massive federal agency that includes organizations like the Food and Drug Administration and the Centers for Disease Control and Prevention. And shortly, O’Neill is likely to join forces formally with Kennedy.

President Donald Trump has nominated O’Neill for the No. 2 position at the agency. The Republican-controlled Senate is likely to confirm Kennedy’s appointment as early as next week, and O’Neill’s nomination is also expected to get the nod this spring.

As O’Neill and Kennedy prepare to take office, Elon Musk and his Department of Government Efficiency loom large in the background. Like O’Neill and Kennedy, Musk has railed against the FDA in the past, arguing at a Trump campaign in the fall that “overregulation kills people.”

The rhetoric from O’Neill and Kennedy has prompted some HHS employees to worry that the duo could work with Elon Musk and DOGE to shake up HHS, just as DOGE has been upending other federal agencies like the U.S. Agency for International Development over the past week.

Along with other federal agencies, HHS employees have already been offered deferred resignation through Musk’s “Fork in the Road” program, and have until February 10 to decide whether to accept. Meanwhile, the White House is reportedly preparing an executive order to cut thousands of jobs from the agency, which currently employs over 80,000 people.

O’Neill’s imminent arrival has sparked enthusiasm among some libertarian-leaning healthcare entrepreneurs and investors who hope he can help unlock a flood of biomedical innovation. Bryan Johnson, the centimillionaire longevity enthusiast, who has known O’Neill for years, is one of those people.

“The nature of drug discovery has dramatically changed, so we’re going to have to upgrade the regulatory environment to match the technological discovery that we’re going to be experiencing,” Johnson told me.

At HHS, O’Neill will have authority over the agency’s daily operations as well as significant influence on research programs and the development of regulations. O’Neill and Kennedy, who squeaked through a key Senate committee vote earlier in the week, will be responsible for allocating healthcare funding, setting food safety standards, regulating drug pricing and managing public communications in the case of a pandemic.

O’Neill, Kennedy and HHS did not respond to requests for comment on this story.

O’Neill arrives at this lofty perch after a career built on reconsidering the balance between personal freedoms and public health. He has also had tours of duty at several Thiel-run organizations, including his time co-founding the Thiel Fellowship.

A staunch libertarian, O’Neill has opposed public health measures such as vaccine mandates and fluoride in drinking water as invasions of personal choice. Meanwhile, he has advocated for measures to accelerate scientific advancement, like automatically approving medications already cleared in Europe and rapidly adopting medical AI technologies. He has applied this mentality to his investment portfolio, backing companies operating on the fringes of modern science, from cryogenics to anti-aging therapeutics.

With O’Neill and Kennedy taking over HHS, a dose of abrupt change is about to hit the American healthcare system, and Silicon Valley’s fingerprints are all over their plans. During Kennedy’s failed presidential campaign, he received hundreds of thousands of dollars from tech heavyweights including Jack Dorsey, David Sacks and Chamath Palihapitiya.

More recently, 8VC partner Sebastian Caliri has been “working closely with many of our respected friends coming in to run HHS,” according to a tweet by 8VC founder Joe Lonsdale. This month, Caliri published a report titled “A Healthcare Excellence Agenda for the New Administration,” which suggests reforming healthcare payment systems to prioritize health outcomes over volume of work, lifting restrictions on healthcare data sharing, and minimizing the role of Institutional Review Boards in clinical trials.

Members of the tech elite like Musk have keen interest in the regulatory process and an obvious reason for wanting to change it in their favor: Musk’s Neuralink, for instance, depends on FDA approvals to advance clinical trials for its brain-computer interfaces. At the same time, Antonio Gracias, a close Musk ally, has reportedly launched a $100 million bid to take over MDMA drug developer Lykos Therapeutics, which is seeking a new FDA approval strategy after the agency rejected its new drug application last summer.

In recent weeks, Musk and DOGE have embarked on a rolling blitzkrieg across Washington with the aim of cutting $2 trillion from the federal budget. HHS, with a $1.8 trillion annual budget, could be a juicy target for their efforts. Kennedy has already said he plans to eliminate 2,200 jobs across the agency.

This week, DOGE reportedly accessed systems at the Centers for Medicare and Medicaid Services and the Centers for Disease Control and Prevention, following up on earlier promises to target fraud and waste within health agencies. DOGE has also cut off payments to Lutheran Family Services, an HHS beneficiary that provides social services to refugees, while the Trump administration attempted to pause all federal grants and loans, causing turmoil in the government-funded medical research community. Meanwhile, all U.S. health agencies have been ordered to pause communications.

Some public health experts are alarmed about the possible downsides of a radically altered approach to governing U.S. healthcare. “You do need to have people who believe in scientific evidence, and that’s a problem when you have someone like Jim O’Neill,” said Diana Zuckerman, president of the nonpartisan National Center for Health Research. She added, “All those investors, don’t they want to be able to use safe and effective medical products for themselves and their families?”

Like Kennedy, O’Neill doesn’t have any formal medical background, a rarity for such high-level appointments within HHS. He does, however, have experience working for the federal government. O’Neill began his career in the Department of Education and then joined HHS under George W. Bush, climbing the ranks to become its principal associate deputy secretary during Bush’s second term. In his role, he had some powers over policy and programming. While O’Neill fought some regulations, like a proposed oversight rule on the use of algorithms in laboratory operations, Bush’s HHS was generally pro-regulation.

Tevi Troy, a former deputy secretary for the department who worked there with O’Neill, emphasized to me how O’Neill’s government background will give him a definite advantage in accomplishing his goals. Troy noted that Kennedy and some other Trump picks for Health and Human Services don’t have the same experience. (That includes surgeon Martin Makary and TV personality Mehmet Oz [aka Dr. Oz], Trump’s picks for running the FDA and the CMS, respectively.) O’Neill’s deeper government experience could mean the real nexus of power and change within the department comes to center on him.

“He’s in a unique position to have a lot of influence,” Troy told me. “I don’t remember a deputy secretary coming in with as much knowledge of the building as him—especially in conjunction with the other members of the top team who have less experience.”

O’Neill nearly became FDA commissioner during Trump’s first administration, but some of his past comments resurfaced, which may have tripped up his nomination. One came from a 2014 biotech conference at which O’Neill told the crowd the FDA should only be required to test new pharmaceuticals for safety, not efficacy. “Let people start using them at their own risk,” he said. “Let’s prove efficacy after they’ve been legalized.”

The other came from a 2009 speech at Thiel’s The Seasteading Institute in which O’Neill proposed changing laws so that organ donors could be compensated, saying, “There are plenty of healthy spare kidneys walking around unused.”

Many of O’Neill’s beliefs mirror Thiel’s—a clear signal of Thiel’s enduring political influence despite his less visible role in the 2024 elections—and the two have a deep relationship.

After Bush left the White House in 2008, O’Neill went to work for Thiel’s Clarium Capital, Mithril Capital and Thiel Foundation. He co-founded the foundation’s biotech investing arm, Breakout Labs, as well as the Thiel Fellowship.

He also shares some of Thiel’s most idiosyncratic interests. While serving on The Seasteading Institute’s board, he became an advocate for charter cities and other forms of politically autonomous special economic zones. Some of these zones, like Próspera in Honduras, have drawn controversy for conducting medical research, including human trials, while bypassing the kind of regulatory oversight they’d be subject to in a country like the U.S.

But O’Neill’s true passion is longevity research. For nearly 12 years, he sat on the board of the SENS Research Foundation, an anti-aging research organization that has received significant Thiel funding. In addition, he served as a board observer for Oisín Biotechnologies, which develops anti-aging gene therapies, and he currently sits on the board of ADvantage Therapeutics, which is working on treatments for neurodegenerative conditions like Alzheimer’s disease. (He will need to recuse himself from the board once he starts his HHS job.)

Fellow advocates in the longevity space have high hopes for O’Neill and the credibility he could lend their field from his perch at the health department. One item on their wish list is for the FDA to classify aging as a disease, which could unlock federal research funding, prompt insurance coverage of anti-aging therapeutics and accelerate medical interventions targeting age-related decline. O’Neill has in fact proposed creating an office of geroscience to redirect funding to anti-aging research.

Christine Peterson, founder of Foresight Institute, a nonprofit dedicated to frontier technology, hopes O’Neill can instill a different approach to drug development within America’s biotech and pharma industries.

“It would be much more cost-effective to say, ‘No, let’s direct our research money on the seven to nine drivers of all chronic disease,’” said Peterson, who first met him decades ago when Thiel began donating to Foresight Institute. “If we can nail even a few of those, we are going to gain so much health span and greatly reduce the money we’re spending for these chronic diseases.”

You’d be hard-pressed to find many stakeholders—from doctors to patients to investors—who are eager to defend the status quo in today’s U.S. healthcare system. America spends more money per capita on healthcare than most other developed nations and yet falls behind many other such countries on metrics like life expectancy, infant mortality and overall population health. The process of getting a new drug to market is slow, expensive and cumbersome, discouraging many would-be entrepreneurs from entering the space.

“Just to enter the process of getting FDA approved, you have to hire multiple consultants,” said Johnson, who has invested in several longevity-focused biotech companies. “It’s too complicated. It’s too arcane. It’s too hard to address. Now, that does not mean that I’m suggesting lower safety. We can maintain the same safety. It’s just extraordinarily hard to try to figure out how to navigate.”

Peterson believes O’Neill’s unusual background will make him a more honest investigator. She said, “It would be really nice if we had folks at the FDA who are not quite so tightly connected to big pharma, because here’s the thing—Jim does biotech advising, but he advises startups,” an approach she believes will lead him to favor disruption over entrenched interests.

The radical change O’Neill wants to bring to HHS could get stress-tested quite soon. Cases of H5N1 avian influenza have soared, and 67 human cases have been confirmed. Just a month ago, the CDC reported the first human death from the flu.

Before leaving office, the outgoing Biden administration awarded $590 million to Moderna, which produced one of the first Covid-19 vaccines, to continue developing a bird flu vaccine.

As recent experience in the U.S. shows, a pandemic tests a country’s ability to communicate public health directives effectively and innovate rapidly—both core responsibilities for HHS. Kennedy and his team’s skepticism of vaccines, masking and social distancing have concerned many public health experts, who worry it could exacerbate a crisis should a new pandemic arise. But if bird flu or a similar virus does make the leap to humans, some biotech leaders are optimistic that O’Neill and the rest of the Trump administration would channel the same energy that led to the rapid development of a Covid-19 vaccine during Trump’s first term.

The Trump world’s tendency to plow through political gridlock could actually be an asset in the case of another outbreak, said Mark Emalfarb, founder and CEO of Dyadic International, a company focused on protein-based biotech, including vaccines.

“Pharma hasn’t really retooled in 35 years,” he said, a problem that affects not only pandemic preparedness but everything from the cost of insulin to the availability of cancer treatments. “They need to get their heads out of the sand because God forbid this thing comes to humans,” he said of the bird flu.

Emalfarb was blunt about the situation. “These pharma companies don’t want to change, and the government doesn’t force them,” he said. “So radical change might actually get the job done that we actually need, because I do believe healthcare is an inalienable right to everybody.”

WSJ : Trump’s Role in TikTok Talks: Dealmaker in Chief

Trump’s Role in TikTok Talks: Dealmaker in Chief
Vance tapped to help oversee negotiations, with companies including Oracle and Amazon signaling interest while TikTok officials hope to avoid a sale

President Trump has been a real-estate tycoon and reality-TV star. Now he’s playing a new role as he oversees TikTok’s fate: investment banker.

In public and behind the scenes, Trump has discussed possible deals to Americanize control of the Chinese-controlled app he once tried to ban, promoting it as a prize asset for tech titans to bid over, a potential target for a planned American sovereign-wealth fund, and a valuable chip in trade negotiations with Beijing.

“GREAT INTEREST IN TIKTOK! Would be wonderful for China, and all concerned,” he posted recently on Truth Social.

Several companies have privately expressed interest in participating in a deal for TikTok or its American operations, including Oracle, Amazon and Microsoft, according to people familiar with the matter. TikTok and its backers, meanwhile, are pushing for solutions that might forego a sale in favor of reviving a plan the company says would wall off American users’ data.

Last week, TikTok Chief Executive Shou Chew met with senior White House officials and offered a proposal that envisioned a joint venture with U.S. investors. That new venture would be headquartered in the U.S. and oversee data security, according to a person familiar with the proposal. Management would be U.S.-based and a board of directors would be a majority U.S. members.

Whether the investors include the U.S. government itself is an open question. Trump on Monday ordered the creation of a sovereign-wealth fund and suggested it could be used to acquire TikTok—something many observers see as far-fetched.

Trump, meantime, has tasked Vice President JD Vance with overseeing negotiations, with the hope Vance’s background in venture capital and Silicon Valley ties can facilitate a deal.

Hope and confusion
Trump’s role and his recent enthusiasm for the app have fueled optimism in TikTok’s camp that it will get a solution to stay in the U.S. long term. It faces an April 5 deadline, although Trump could extend that.

But his maneuvers also have ignited confusion over what any such deal would look like and who would be involved.

“The president has made it clear he wants TikTok to continue. He has determined as dealmaker-in-chief that he will find a path forward,” said a person close to the company. But the person added that talk of a sovereign fund had muddied the conversation. “We don’t even have a sovereign-wealth fund yet…We don’t have time to wait.”

Trump’s executive order on inauguration day set a 75-day delay in enforcement of a bipartisan law passed last year requiring TikTok to shed its Chinese ownership or close in the U.S. The law’s proponents argue that China’s government could pressure its Chinese parent, ByteDance, to use TikTok to surveil or propagandize Americans—a claim TikTok has disputed.

Trump has indicated a range of potential outcomes, at times suggesting a U.S. company or even the U.S. itself should own TikTok and at other times indicating U.S. investors could just increase their stakes in its parent company.

ByteDance has said that it is already roughly 60% owned by global institutional investors, many of which are big U.S. financial firms—including BlackRock, General Atlantic and Susquehanna International Group, co-founded by Republican megadonor Jeff Yass.

After trying to ban TikTok in 2020, Trump embraced it during the latest presidential campaign, in part because he came to view it as a powerful political tool. He was partially persuaded by Kellyanne Conway, a senior adviser in his first term who has worked on behalf of TikTok allies to advocate for it. She told him he had more supporters on the platform than Kamala Harris or Joe Biden—and that many young people made videos in his favor. She also conducted polling that showed Americans didn’t see banning TikTok as the best way to counter China.

In a December meeting with TikTok’s CEO at the president’s Mar-a-Lago estate in Florida, Trump phoned his son, Barron, and bragged about his statistics on the platform and told Barron that the company’s CEO had flown in to see him. “I’m going to save TikTok,” he told an adviser after that meeting.

A deal for TikTok is now expected to be part of broader talks on trade and tariffs that play out over the coming month. Trump had been expected to speak this week with Chinese leader Xi Jinping but that hadn’t happened as of late Friday. White House staff have been engaging with Chinese counterparts, according to people briefed on those discussions, as they look to define parameters for talks. The two discussed TikTok, among other issues, in a Jan. 17 call.

In Beijing, Chinese officials have told ByteDance that they were interested in reaching a deal with the U.S. to resolve the dispute and were open to the option of a U.S. sovereign-wealth fund investing in TikTok for a significant stake, people familiar with the matter said. They said they hoped that a deal could help Beijing with negotiations for tariffs and other issues.

Talking up a bidding war
Trump has encouraged potential suitors in an apparent effort to stoke maximum interest. “I like bidding wars because you make your best deal. So if there’s a bidding war, that’s a good thing,” he told reporters while traveling on Air Force One late last month.

On his second day in office, he told a room of reporters he would support Elon Musk buying the app. Then he added that he would also welcome Larry Ellison, Oracle’s chairman, buying it. Asked on the Air Force One trip whether Microsoft could play a role, Trump said: “I would say yes. A lot of interest in TikTok.”

A few days later he floated the idea that the U.S. could buy half of TikTok, before this week saying he wants to establish the U.S. sovereign-wealth fund.

Trump has said that his power to decide TikTok’s fate should entitle the U.S. government to a sizable stake in whatever entity emerges. “If I sign, then somebody’s gonna buy it, pay a lot of money, have a lot of jobs, keep a platform open and have it be very secure. If I don’t sign, then it closes,” he said.

His role has so far left actual investment banks on the sidelines of what could be one of the biggest deals in recent history, boxing out the bankers who typically provide more clarity to potential bidders.

“It is unprecedented for a president to introduce himself into a sale of a private asset,” said Darrell West, a senior fellow at the Brookings Institution who specializes in technology. “Typically presidents trust the market to do those things.”

Trump tried something similar in his first term. After initiating an effort to ban TikTok, he blessed a tentative deal to save it in which Oracle and Walmart would have taken stakes in a U.S.-based TikTok. The agreement never materialized.

Investor consortia are cropping up with vastly different ideas on how a deal could be structured. One led by billionaire Frank McCourt—which counts as a partner Kevin O’Leary, a star on “Shark Tank” known as Mr. Wonderful—proposes a bid worth more than $20 billion.

Another investor group pitched the idea of investing in a special purpose vehicle that would buy TikTok and invest in Musk’s artificial intelligence startup, xAI, with the idea that it would run TikTok’s algorithm, according to someone familiar with the group.

ByteDance and its investors aim to convince Trump that an earlier plan by the company to transfer U.S. user data to Oracle’s cloud satisfies security concerns, people close to the company said.

TikTok started crafting that initiative, dubbed Project Texas, during Trump’s previous administration in an effort to convince the Committee on Foreign Investment in the U.S. that the company could satisfy American national security concerns without separating from its Chinese parent company.

TikTok executives believed Trump would have approved Project Texas as the solution to lawmakers’ concerns had he won re-election in 2020. Instead, negotiations with Washington about the initiative fell apart under President Biden.

WSJ : Sudan’s Army Close to Retaking its Own Capital From Rebels

Sudan’s Army Close to Retaking its Own Capital From Rebels
Offensive reverses months of setbacks for military government while rebel discipline shows signs of fraying

After a lightning offensive, Sudan’s military is close to regaining control of the country’s capital, Khartoum, for the first time since a rebel general plunged the East African nation into civil war nearly two years ago.

Government troops have seized multiple neighborhoods in the ruins of central Khartoum in recent days and cornered rebels of the Rapid Support Forces in their most significant remaining stronghold, the Republican Palace, the presidential residence.

At its core, the war is a power struggle between the country’s de facto president, Lt. Gen. Abdel Fattah al-Burhan, and his former No. 2, RSF commander Lt. Gen. Mohamed Hamdan Dagalo, that has left tens of thousands dead, and over 12 million displaced from their homes, according to the United Nations.

“We are ready to remove the militias from every corner of Sudan,” Burhan, dressed in fatigues, told cheering soldiers after retaking the military headquarters late last month.

The U.S. in January accused Dagalo of genocide because his mostly Arab paramilitary force has killed thousands of Black Sudanese in the country’s western Darfur region. It was the second alleged genocide committed by the RSF, whose precursor, the Janjaweed militias, killed some 200,000 Black Sudanese in the early 2000s.

The U.S. also accused Burhan of war crimes, mostly because his warplanes have bombed civilian populations.

The RSF seized control of most of the capital soon after the war started in April 2023, setting up checkpoints on streets and posting hundreds of snipers on rooftops in a city largely reduced to rubble by the fighting.

The Sudanese army struggled to regain its footing.

In recent weeks, however, the army has recaptured the country’s second-largest city, Wad Madani, and taken a string of towns close to the South Sudan border.

Last month, government forces retook the country’s largest oil refinery, located north of Khartoum, and broke a siege of the army’s main Khartoum base.

“The RSF has seen their supply lines stretched thin after two years of war,” said Cameron Hudson, a former State Department official now at the Center for Strategic and International Studies in Washington, D.C. “They are struggling with desertions and [with] Sudan’s military that is for the first time at its full strength.”

Analysts say the army’s resurgence has been powered by new weapons from Iran, an influx of recruits and the backing of volunteer militias.

Last week, Doctors Without Borders, the French medical charity, said it was contending with a mass influx of wounded in Khartoum and Darfur. Some 58 people were killed and more than 150 others injured after the RSF shelled a busy market in Khartoum’s twin city of Omdurman, according to the Sudanese health ministry.

“The violence continues to ruin lives, making it harder for people to access healthcare and putting healthcare workers at risk,” said Ozan Agbas, the Doctors Without Borders emergency manager.

The battle for control of the Republican Palace could prove decisive for the RSF, whose discipline has begun to show signs of fraying. Losing the palace and adjacent neighborhoods south of the Blue Nile would further disrupt the rebels’ ability to receive supplies from western Sudan, their home base.

RSF commander Dagalo, widely known by his nickname, Hemedti, hasn’t appeared in public for months, raising speculation about his whereabouts. His elder brother, Abdul Rahim Hamdan Dagalo, the rebel group’s deputy commander, has assumed command.

His absence has demoralized his fighters, hundreds of whom have retreated from the capital toward their homes in Darfur, according to Sudanese officials and activists.

“Most RSF fighters feel Hemedti has abandoned them,” a former RSF fighter living in exile in Uganda told The Wall Street Journal.

An RSF spokesman said Saturday that reports of the army’s victories were exaggerated. “The RSF still controls most of the capital Khartoum,” he said.

Should the RSF lose more ground in the city, some analysts expect the fighting to move to Darfur, raising the risk of further violence aimed at the region’s Black population.

The war’s shifting fortunes are also likely to ripple through the region, where other governments have used the conflict to advance their own interests in Sudan, which is situated along key Red Sea trade routes.

For instance, the United Arab Emirates armed the rebels as recently as 2023, The Wall Street Journal has reported.

An Emirati official denied the Gulf state is taking sides in the conflict. “We continue to urgently call for an immediate cease-fire and a peaceful resolution to this man-made conflict,” the official said. “In this regard, the U.A.E. has already made absolutely clear that it is not providing any support or supplies to either of two belligerent warring parties in Sudan.”

For its part, the Sudanese military has acknowledged receiving drones from Iran. The army’s “increasing reliance on Iran for weapons and financing presents the danger of Sudan becoming another Iranian proxy,” said Donald E. Booth, a former U.S. envoy to Sudan now with the Wilson Center, a Washington think tank.

Barron's : The Mag 7’s Strength Just Became a Weakness

The Mag 7’s Strength Just Became a Weakness

Not that long ago, the cloud was Big Tech’s greatest strength. Now it might be its kryptonite.

Amazon.com, Microsoft, and Google parent Alphabet have recently reported generally impressive numbers, but their stocks all fell largely because of very slight cloud misses. Meanwhile, Meta Platforms stock soared. The difference? Meta has no public cloud to rain on the parade.

The earnings reports made clear that the breathtaking pace of capital expenditures for AI data centers will continue in 2025. That has raised the bar across Big Tech, especially for the cloud, where a growth slowdown has long been inevitable.

Amazon, which pioneered the public cloud category with Amazon Web Services, sees roughly 20% annualized growth from its AWS cloud business, down from over 50% five years ago. Now Google and Microsoft may be experiencing a similar trend. Google Cloud Platform and Microsoft Azure are still growing 30% annually, but the latest earnings suggest that growth may be decelerating. That reality pressured shares over the past two weeks, and the pain may not be over.

Cloud softness feels particularly acute, with tech companies spending gobs of money to build out artificial-intelligence capabilities.

During its fourth-quarter earnings call on Thursday, Amazon projected around $105 billion on capital expenditures in 2025, a 27% rise from 2024, which was 57% over 2023. Microsoft has guided to $80 billion in capex for its fiscal 2025, ending June, up 80% from 2024, which was up 58% from the year before. Google estimated capex of $75 billion in 2025, up 43% from 2023, which was up 63% from 2022.

Meta Platforms is moving into the same neighborhood, with a forecast of $60 billion to $65 billion of capex in 2025, up 68% at the midpoint from 2024, which was up by 37% from the year before. Meta differs from the others in that it doesn’t have a cloud unit that rents out servers; these new data centers are all for Meta’s use.

The capex numbers are eye-watering, and investors want to know how long it will take for revenue and profit to begin flowing from them. Rolling out sales-generating AI services will take time, but the immediate place new AI revenue has emerged is in accelerated growth rates in the cloud units of Amazon, Microsoft, and Google. But this latest quarter raised some doubts even there.

Amazon’s fourth-quarter results were a mixed bag —and it wasn’t just because of the cloud. Overall, the company beat on sales and profitability, but that came with narrow misses in its two fastest-growing segments, AWS and advertising. Its overall first-quarter guidance was also disappointing. Amazon shares were down around 4% on Friday.

Microsoft posted solid consolidated numbers, beating analysts’ expectations on both revenue and earnings per share. Yet the stock fell 6% the next day. The culprit was annual growth in its Azure cloud unit of 31%—spectacular, but not spectacular enough for investors whose patience is beginning to run thin. That growth rate was at the low end of Microsoft’s guidance, and represented deceleration.

“Growing a business as big as Azure 31% is pretty amazing but it has decelerated from growth of 33%, 34%, and 35% in the prior three quarters—and was widely expected to accelerate in the March and June 2025 quarters,” Melius analyst Ben Reitzes wrote in a note to clients. “It doesn’t look like that is happening.”

Microsoft attributed the Azure miss to continued supply constraints for Nvidia AI accelerators, and to allocating too many sales teams to AI at the expense of traditional servers that still run things like websites, online games, and web apps.

Cue the selloff.

Alphabet also beat Wall Street expectations for earnings per share, but fell short on revenue by a hair. Contributing to that was cloud revenue of $12 billion, when Wall Street expected $12.2 billion. Google Cloud grew revenue by 30% over 2023, but analysts wanted 33%. Both numbers marked a deceleration from the previous quarter, when growth was 35%.

Google attributed the miss to the same supply issues Microsoft had. Here too, investors didn’t seem interested in nuance. Google shares declined 7% on the news.

Two Big Tech companies stood out. With no cloud unit to report, Meta’s results simply couldn’t disappoint in the same way as some of its rivals. Meta had nothing to draw attention away from its advertising growth of 21% in the fourth quarter. Meta stock rose 1.6% the day after reporting.

Among the Big Tech group, Apple is alone in sitting out the AI arms race, spending only $10 billion on 2024 capex, and not guiding to any sort of sudden rise. Apple relies on rented computing in other companies’ clouds, its own chips, and small on-device AI models. In a post-DeepSeek world, where models could be commoditized, Apple may have settled on the right strategy.

“The capex surge at Alphabet highlights perhaps why Apple is able to garner such a relatively high multiple in the Mag 7 these days as it may prove to be the business model that benefits most from lower model costs—generating more free cash flow than net income,” Reitzes wrote.

Apple shares were down just 0.7% after its report.

FT : House of Gucci needs a spring clean

House of Gucci needs a spring clean
It would be wrong to bet against the brand reinventing itself again

“In difficult times, fashion is always outrageous,” the eccentric Italian designer Elsa Schiaparelli once said. So it is proving at one of luxury’s most storied brands. Gucci, part of the Kering corporate empire, parted ways with its creative director Sabato De Sarno this week after slumping sales and multiple profit warnings.

Luxury houses are used to highs and lows. The issue here is that Kering lacks many of the soft layers that pad its peers. Gucci — which accounts for two-thirds of operating profit — caters to the affluent more than the astronomically wealthy, which leaves it exposed to the economic cycle. Its over-the-top brand heritage, too, can be in or out of favour. It doesn’t sit well with the current “quiet luxury” movement.

In less status-focused industries, managers try to cut costs when sales dip. That’s harder for companies like Gucci, which have an image to maintain. Fully three-quarters of Gucci’s first-half revenue decline passed straight through to its operating profit.


Alongside a more cyclical and concentrated business mix, Kering has also amassed more debt. Its peers may entice customers to splurge, but they keep a close eye on their own finances. Richemont and Hermes hold cash. LVMH’s net debt is roughly equivalent to its forecast ebitda. On S&P Capital IQ numbers, the Gucci owner’s indebtedness including capitalised leases at the end of June 2024 was three times that.

It’s not like François-Henri Pinault has nothing to show for this. Kering has acquired 30 per cent of Valentino, perfumer Creed and an eye-popping €1.3bn building in Milan’s Via Montenapoleone. But there may be further calls on its resources. The 70 per cent of Valentino Kering does not already own is governed by put and call options exercisable from May 2026 to August 2028 and is in the books at €4bn.


Investors will no doubt be scouring Kering’s annual results next week for guidance on what happens next. The group may hope that a rising tide will help it float above such matters. After all, the luxury cycle seems to be turning. Yet with no designer at the helm, Kering risks missing out on this particular wave. The upshot is that Pinault may well come under pressure to offload real estate stakes — as it did with some Parisian properties in January — and embark on some serious cost cutting.

The House of Gucci is no stranger to drama — it has endured M&A battles, departures of iconic designers and bitter feuds. It would be wrong to bet against the brand reinventing itself again. But unless sales growth materialises, its next new look will need a slimmer fit.

Barron's : This Garbage Stock Isn’t Garbage. Why It’s Time to Buy GFL Now.

This Garbage Stock Isn’t Garbage. Why It’s Time to Buy GFL Now.
Waste hauler GFL Environmental has trailed its peers. With a transformational deal in the works, that’s about to change.

Waste hauler GFL Environmental is a catalyst-rich stock in a profitable industry, interesting in any environment and especially appealing in this one. Its plan for the future sets it up for success, and it should see little impact from artificial intelligence, politics, or the broader market. That’s a rare and powerful combination that should catch investor attention.

Waste hauler GFL Environmental is a catalyst-rich stock in a profitable industry, interesting in any environment and especially appealing in this one. Its plan for the future sets it up for success, and it should see little impact from artificial intelligence, politics, or the broader market. That’s a rare and powerful combination that should catch investor attention.
Appealing might not be the first term that comes to mind with garbage, but it should apply to waste-management stocks. Historically, they are stable growers that do well in most market environments because they serve an essential, universal need. The way our detritus is being discarded is gradually becoming more complicated, as recycling and composting gain in popularity. The companies have also learned how to capture renewable natural gas from their landfills, offering another potential opportunity.

More waste volume and complex handling boil down to more growth—and higher stock prices—for garbage companies. Waste Connections, Waste Management, Republic Services, and Casella Waste Systems have returned an average of 59%, including reinvested dividends, over the past three years, beating the S&P 500 index’s 40% total return over the same period. GFL, however, has returned just 39%, the only one of the group to underperform the benchmark.

The company, which is based in Vaughan, Ontario, also has offices in Raleigh, N.C., and conducts about 70% of its waste-management business in the U.S. It’s unlikely to be affected by tariffs.

One big issue: GFL is carrying a lot of debt. It has some $6.6 billion due between 2025 and 2032, about four times 2025’s earnings before interest, taxes, depreciation, and amortization, or Ebitda. Industrial companies in the S&P 500 typically operate with a ratio of less than two times, and other waste players are all between two and three times.

High debt is a legacy of GFL’s ownership and strategy. Before its IPO in 2020, GFL was owned by private-equity firms led by BC Partners, which were happy to see founder and CEO Patrick Dovigi expand the company by rolling up other waste haulers. BC Partners still holds about 24% of the shares outstanding, which creates another problem for investors: Private-equity players typically don’t hold publicly traded stocks for the long haul—and no one wants to buy a stock right before a large holder liquidates a position.

All that is changing. In January, the company signed an agreement to sell its hazardous-waste-handling environmental services business for an enterprise value of $5.6 billion to Apollo and BC Partners. GFL will receive some $4.3 billion in cash proceeds and retain a minority equity stake worth $1.2 billion. With the cash, GFL will repay about $2.7 billion in debt and repurchase $1.6 billion of shares—likely from BC Partners.

“They’re selling their environmental services assets at a premium to where the current [company] is trading at despite the fact that solid waste is a better business…higher margin, lower capital intensity, better free-cash-flow conversion,” says Brandon Geisler, a portfolio manager at Fred Alger Management.

The environmental services business is going for about 16 to 18 times estimated 2025 Ebitda—depending on how investors want to allocate overhead. It leaves GFL’s solid-waste business trading for about 13 times. Peers trade for an average of about 16 times.

There is little reason for a discount after the deal is done. GFL’s debt will be reduced, and the balance sheet will be investment-grade. Less stock will be in the hands of private-equity players. What’s more, GFL will still have the $1.2 billion stake in the environmental services business run by Apollo and BC Partners.

Looking ahead, Wall Street projects about 6% sales growth for the solid-waste business over the coming three years, with margins expanding another two percentage points.

“We see a path for the company’s margins to expand at a greater rate compared with the Big 3,” writes CIBC analyst Kevin Chiang, referring to Republic Services, Waste Connections, and Waste Management. He projects about 1.5 percentage points of margin improvement in 2025. Chiang rates shares Buy, with a $53 target price target.

With a better balance sheet and a renewed focus on solid waste, GFL’s growth can beat the current targets, a strong catalyst for higher multiples after investors evaluate the remaining business. Seeing the valuation gap vanish after the sales would push shares into the mid-$50s from Tuesday’s close of $44.18.

GFL’s remaining stake in the environmental services business and a 45% stake in road paver Green Infrastructure Partners are worth roughly $4 to $5 a share. That’s nothing to throw away.

So why the discount? For starters, it’s a complicated situation and investors are waiting. The debt paydown and stock repurchase are coming, and the deal is closing. The company is hosting an investor event in February to help demystify the stock and the future.

Post-deal, “we’re going to look like everybody else,” says Dovigi. With the business and balance sheet restructured, he’s focused on doubling free cash flow over the coming four or five years—similar to GFL’s performance between 2020 and 2024.

If he’s successful, the coming catalysts will position GFL to trade at a premium in the waste space—and as a winning stock.

TechCrunch : Tesla’s Dojo, a timeline

Tesla’s Dojo, a timeline

Elon Musk doesn’t want Tesla to be just an automaker. He wants Tesla to be an AI company, one that’s figured out how to make cars drive themselves.

Crucial to that mission is Dojo, Tesla’s custom-built supercomputer designed to train its Full Self-Driving (FSD) neural networks. FSD isn’t actually fully self-driving; it can perform some automated driving tasks, but still requires an attentive human behind the wheel. But Tesla thinks with more data, more compute power and more training, it can cross the threshold from almost self-driving to full self-driving.

And that’s where Dojo comes in.

Musk has been teasing Dojo for some time, but the executive ramped up discussions about the supercomputer throughout 2024. Now that we’re in 2025, another supercomputer called Cortex has entered the chat, but Dojo’s importance to Tesla might still be existential — with EV sales slumping, investors want assurances that Tesla can achieve autonomy. Below is a timeline of Dojo mentions and promises.

2019
First mentions of Dojo

April 22 – At Tesla’s Autonomy Day, the automaker had its AI team onstage to talk about Autopilot and Full Self-Driving, and the AI powering them both. The company shares information about Tesla’s custom-built chips that are designed specifically for neural networks and self-driving cars.

During the event, Musk teases Dojo, revealing that it’s a supercomputer for training AI. He also notes that all Tesla cars being produced at the time would have all hardware necessary for full self-driving and only needed a software update.

2020
Musk begins the Dojo roadshow

Feb 2 – Musk says Tesla will soon have more than a million connected vehicles worldwide with sensors and compute needed for full self-driving — and touts Dojo’s capabilities.

“Dojo, our training supercomputer, will be able to process vast amounts of video training data & efficiently run hyperspace arrays with a vast number of parameters, plenty of memory & ultra-high bandwidth between cores. More on this later.”

August 14 – Musk reiterates Tesla’s plan to develop a neural network training computer called Dojo “to process truly vast amounts of video data,” calling it “a beast.” He also says the first version of Dojo is “about a year away,” which would put its launch date somewhere around August 2021.

December 31 – Elon says Dojo isn’t needed, but it will make self-driving better. “It isn’t enough to be safer than human drivers, Autopilot ultimately needs to be more than 10 times safer than human drivers.”

2021
Tesla makes Dojo official

August 19 – The automaker officially announces Dojo at Tesla’s first AI Day, an event meant to attract engineers to Tesla’s AI team. Tesla also introduces its D1 chip, which the automaker says it will use — alongside Nvidia’s GPU — to power the Dojo supercomputer. Tesla notes its AI cluster will house 3,000 D1 chips.

October 12 – Tesla releases a Dojo Technology whitepaper, “a guide to Tesla’s configurable floating point formats & arithmetic.” The whitepaper outlines a technical standard for a new type of binary floating-point arithmetic that’s used in deep learning neural networks and can be implemented “entirely in software, entirely in hardware, or in any combination of software and hardware.”

2022
Tesla reveals Dojo progress
August 12 – Musk says Tesla will “phase in Dojo. Won’t need to buy as many incremental GPUs next year.”

September 30 – At Tesla’s second AI Day, the company reveals that it has installed the first Dojo cabinet, testing 2.2 megawatts of load testing. Tesla says it was building one tile per day (which is made up of 25 D1 chips). Tesla demos Dojo onstage running a Stable Diffusion model to create an AI-generated image of a “Cybertruck on Mars.”

Importantly, the company sets a target date of a full Exapod cluster to be completed by Q1 2023, and says it plans to build a total of seven Exapods in Palo Alto.

2023
A ‘long-shot bet‘

April 19 – Musk tells investors during Tesla’s first-quarter earnings that Dojo “has the potential for an order of magnitude improvement in the cost of training,” and also “has the potential to become a sellable service that we would offer to other companies in the same way that Amazon Web Services offers web services.”

Musk also notes that he’d “look at Dojo as kind of a long-shot bet,” but a “bet worth making.”

June 21 – The Tesla AI X account posts that the company’s neural networks are already in customer vehicles. The thread includes a graph with a timeline of Tesla’s current and projected compute power, which places the start of Dojo production at July 2023, although it’s not clear if this refers to the D1 chips or the supercomputer itself. Musk says that same day that Dojo was already online and running tasks at Tesla data centers.

The company also projects that Tesla’s compute will be the top five in the entire world by around February 2024 (there are no indications this was successful) and that Tesla would reach 100 exaflops by October 2024.

July 19 – Tesla notes in its second-quarter earnings report it has started production of Dojo. Musk also says Tesla plans to spend more than $1 billion on Dojo through 2024.

September 6 – Musk posts on X that Tesla is limited by AI training compute, but that Nvidia and Dojo will fix that. He says managing the data from the roughly 160 billion frames of video Tesla gets from its cars per day is extremely difficult.

2024
Plans to scale

January 24 – During Tesla’s fourth-quarter and full-year earnings call, Musk acknowledges again that Dojo is a high-risk, high-reward project. He also says that Tesla was pursuing “the dual path of Nvidia and Dojo,” that “Dojo is working” and is “doing training jobs.” He notes Tesla is scaling it up and has “plans for Dojo 1.5, Dojo 2, Dojo 3 and whatnot.”

January 26 – Tesla announced plans to spend $500 million to build a Dojo supercomputer in Buffalo. Musk then downplays the investment somewhat, posting on X that while $500 million is a large sum, it’s “only equivalent to a 10k H100 system from Nvidia. Tesla will spend more than that on Nvidia hardware this year. The table stakes for being competitive in AI are at least several billion dollars per year at this point.”

April 30 – At TSMC’s North American Technology Symposium, the company says Dojo’s next-generation training tile — the D2, which puts the entire Dojo tile onto a single silicon wafer, rather than connecting 25 chips to make one tile — is already in production, according to IEEE Spectrum.

May 20 – Musk notes that the rear portion of the Giga Texas factory extension will include the construction of “a super dense, water-cooled supercomputer cluster.”

June 4 – A CNBC report reveals Musk diverted thousands of Nvidia chips reserved for Tesla to X and xAI. After initially saying the report was false, Musk posts on X that Tesla didn’t have a location to send the Nvidia chips to turn them on, due to the continued construction on the south extension of Giga Texas, “so they would have just sat in a warehouse.” He noted the extension will “house 50k H100s for FSD training.”

He also posts:

“Of the roughly $10B in AI-related expenditures I said Tesla would make this year, about half is internal, primarily the Tesla-designed AI inference computer and sensors present in all of our cars, plus Dojo. For building the AI training superclusters, NVidia hardware is about 2/3 of the cost. My current best guess for Nvidia purchases by Tesla are $3B to $4B this year.”

July 1 – Musk reveals on X that current Tesla vehicles may not have the right hardware for the company’s next-gen AI model. He says that the roughly 5x increase in parameter count with the next-gen AI “is very difficult to achieve without upgrading the vehicle inference computer.”

Nvidia supply challenges
July 23 – During Tesla’s second-quarter earnings call, Musk says demand for Nvidia hardware is “so high that it’s often difficult to get the GPUs.”

“I think this therefore requires that we put a lot more effort on Dojo in order to ensure that we’ve got the training capability that we need,” Musk says. “And we do see a path to being competitive with Nvidia with Dojo.”

A graph in Tesla’s investor deck predicts that Tesla AI training capacity will ramp to roughly 90,000 H100 equivalent GPUs by the end of 2024, up from around 40,000 in June. Later that day on X, Musk posts that Dojo 1 will have “roughly 8k H100-equivalent of training online by end of year.” He also posts photos of the supercomputer, which appears to use the same fridge-like stainless steel exterior as Tesla’s Cybertrucks.

From Dojo to Cortex
July 30 – AI5 is ~18 months away from high-volume production, Musk says in a reply to a post from someone claiming to start a club of “Tesla HW4/AI4 owners angry about getting left behind when AI5 comes out.”

August 3 – Musk posts on X that he did a walkthrough of “the Tesla supercompute cluster at Giga Texas (aka Cortex).” He notes that it would be made roughly of 100,000 H100/H200 Nvidia GPUs with “massive storage for video training of FSD & Optimus.”

August 26 – Musk posts on X a video of Cortex, which he refers to as “the giant new AI training supercluster being built at Tesla HQ in Austin to solve real-world AI.”

2025
No updates on Dojo in 2025
January 29 – Tesla’s Q4 and full-year 2024 earnings call included no mention of Dojo. Cortex, Tesla’s new AI training supercluster at the Austin gigafactory, did make an appearance, however. Tesla noted in its shareholder deck that it completed the deployment of Cortex, which is made up of roughly 50,000 H100 Nvidia GPUs.

“Cortex helped enable V13 of FSD (Supervised), which boasts major improvements in safety and comfort thanks to 4.2x increase in data, higher resolution video inputs … among other enhancements,” according to the letter.

During the call, CFO Vaibhav Taneja noted that Tesla accelerated the buildout of Cortex to speed up the rollout of FSD V13. He said that accumulated AI-related capital expenditures, including infrastructure, “so far has been approximately $5 billion.” In 2025, Taneja said he expects capex to be flat as it relates to AI.

CrunchBase : The Week’s Biggest Funding Rounds: BlinkRx, Tidal Vision Lead Anoth

The Week’s Biggest Funding Rounds: BlinkRx, Tidal Vision Lead Another Slow week

For the past couple of weeks large rounds have barely trickled in, and this week continues that trend. Just three deals hit $100 million or more as investors seem to have pulled back on big rounds early in the year.

1. (tied) BlinkRx, $140M, pharmaceutical: The biggest raise this week came from the pharmaceutical and prescription industry. BlinkRx, a prescription access platform, raised a $140 million Series D led by 1789 Capital. Donald Trump Jr., a partner at the investment firm, was named to the New York-based startup’s board. BlinkRx promises price transparency on drugs and delivery. Founded in 2014, the company has raised $315 million, per Crunchbase.

1. (tied) Tidal Vision, $140M, biotech: Biotech firm Tidal Vision tied for the biggest round this week, raising a $140 million Series B financing from several investors including Eni Next. The Bellingham, Washington-based company creates scalable biomolecular solutions for critical industries — including water treatment, agriculture and material science. Tidal is building new infrastructure in Europe, Texas and Ohio and will use the new cash to accelerate research and development in chitosan — a nontoxic biopolymer — and adjacent technologies. Founded in 2015, the company has raised nearly $224 million, per Crunchbase.

3. Semgrep, $100M, cybersecurity: Although cybersecurity venture funding bounced back some last year, it ended 2024 being relatively stagnant. This year, not much has changed, but this week did see the biggest round of the year so far for the industry. Application security startup Semgrep raised a $100 million Series D funding led by Menlo Ventures. Lightspeed Venture Partners led the company’s $53 million Series C in 2023. Founded in 2017, Semgrep has raised $204 million, according to the company. Semgrep’s autonomous code security platform allows developers and security engineers to create guardrails that proactively secure application development.

4. (tied) Hidden Level, $65M, defense: Another day and another big round for a defense tech firm. Hidden Level raised a $65 million Series C led by DFJ Growth. The Syracuse, New York-based startup said the new cash comes six months after it raised a previously unannounced $35 million Series B. Hidden Level’s passive radar systems allows users to detect and track drones and other objects in the air — a more common tactic in modern warfare. The passive radar systems cannot identify approaching aircraft without being detected. The company has contracts this year to support deployments for the U.S. Army, U.S. Air Force, U.S. Africa Command, U.S. Indo-Pacific Command, U.S. Central Command, U.S. Northern Command and other federal, state and local agencies. Hidden level is just the latest startup to collect a big check — or two — from VCs interested in defense tech. Just last week, El Segundo, California-based Castelion, a defense manufacturer developing long-range hypersonic strike weapons, raised a $100 million Series A in a mix of debt and equity. Lightspeed Venture Partners led the equity portion.

4. (tied) Nextworld, $65M, enterprise software: Denver-based Nextworld, a provider of enterprise platforms that enable businesses to create tailored software solutions, closed a $65 million Series F led by the McVaney Investment Partnership. The company plans to grow strategic partnerships and accelerate its research and development capabilities. Founded in 2016, this is the first announced round, per Crunchbase.

6. (tied) Berry Street, $50M, healthcare: New York-based Berry Street, a nutrition counseling platform, raised $50 million. No lead investors were named, but investors included Northzone, Sofina and others. Founded in 2022, the company has raised $59 million, per Crunchbase.

6. (tied) Fay, $50M, healthcare: Staying in the same nutritional sector, San Francisco-based Fay, a digital nutritional therapy platform, locked up a $50 million Series B led by Goldman Sachs at a $500 million valuation. Founded in 2022, Fay says it has raised $75 million.

8. (tied) 75F, $45M, smart building: Minneapolis-based 75F, a developer of AI-driven commercial HVAC automation, announced a $45 million Series B funding round led by Accurant International’s Net Zero Alliance. Founded in 2012, the company has raised nearly $75 million, per Crunchbase.

8. (tied) MagicSchool AI, $45M, edtech: Denver-based MagicSchool, an AI platform for education, raised a $45 million Series B led by Valor Equity Partners. Founded in 2023, the company has raised $62 million, per Crunchbase.

10. 7AI, $36M, cybersecurity: Boston-based 7AI, an agentic security platform, raised a $36 million seed. Investors included the likes of Greylock Partners and Spark Capital. Founded in 2024, this is the company’s first raise, per Crunchbase.

Big global deals
The largest deal of the week came from our neighbors to the north.
  • Toronto-based StackAdapt, a multichannel programmatic advertising platform that uses AI and automation in its software to enhance capabilities and user experience, raised a $235 million growth round led by Teachers’ Venture Growth.