TheChek News Flow

DAX:

MDAX:

SDAX:

Why is AI struggling to discover new drugs?

A generation of start-ups have failed to live up to the hype. Executives are now betting that more powerful tools will crack the complexities of human biology

In the mid-2010s, a spate of start-ups hoping to transform the laborious process of finding new drugs launched with big promises. Artificial intelligence would dramatically reduce the time it took to discover new medicines and cut the average of $2bn it takes to develop a drug. 

The emerging businesses attracted the attention of Big Pharma companies such as Bristol Myers Squibb and Sanofi, which signed deals worth billions of dollars pending the drugs’ eventual approval. Press releases boasted of “breakthrough productivity gains” and “groundbreaking research collaborations”.

But now, sceptics are asking: where are the drugs? It has been longer than the average 10 years that it takes to discover and develop a medicine, yet there are few AI-discovered candidates in late-stage clinical trials, and not one has been approved. Despite pledging to cut the industry’s high failure rate, many of the companies’ initial studies flopped. 

Some start-ups have struggled financially, launching in a period where investors have pulled money from the biotech sector in general. BenevolentAI, a British company that attracted lots of early excitement, saw its shares fall more than 99 per cent before it delisted in March, merging with a Japanese company. US-based Recursion snapped up rival Exscientia cheaply last year, paying $688mn — only $180mn more than the cash on its balance sheet and far less than the $2.9bn valuation it went public at three years before.

Alex Zhavoronkov, chief executive of AI for drug discovery company Insilico, says companies have been under pressure to prove their big claims about transforming drug discovery, by showing they had actual drugs. “In order to claim that you have a golden goose, you need to make sure that you have laid a few golden eggs. And if you don’t have golden eggs, your golden goose is depreciating very, very quickly,” he says. 

Daphne Koller, chief executive of another, similarly named, AI for drug discovery start-up, Insitro, says fundamentally we are trying to fix something we do not understand because of the complexity of human biology. “I used to say we were the industry with the highest failure rates of anything but space exploration. And then space exploration started to work,” she says. 

The idea of applying AI to drug discovery has been so attractive because investors see the pharmaceutical sector as a key area where slow and expensive processes are ripe for disruption. Venture capitalists poured money into companies trying to use AI to discover new drugs, seeing the industry as a promising field, especially as ageing populations increase medical costs around the world. Funding for AI drug discovery companies increased from $30mn in 2013 to a peak of $1.8bn in 2021, according to data from PitchBook.

The meteoric rise of generative AI since the launch of ChatGPT in late 2022 has kick-started a new boom in using powerful AI technologies to design drugs.

AI tools are already showing potential in other scientific fields, such as extreme weather prediction. Investors are being lured back in by hope that a fresh crop of companies, technological breakthroughs and new ways to collect and understand biological data could finally lay a golden egg.

The bet behind this new generation of companies is that AI can still revolutionise the drug discovery process, but that the initial tools being used were not powerful enough. As a result, supporters say, it is too soon to rush to judgment.

Yet so far, the problem has proved beyond the algorithms. We know surprisingly little about our own biology. There are many mysteries in how our cells interact and challenges in measuring our body’s most crucial processes, starving models of the data they need to make more rapid progress.

Darren Green, a veteran chemist who spent more than 30 years at GSK, says drug discovery is “probably the hardest thing mankind tries to do”. “We get these great new tools, which is fantastic. And then you just find another problem,” he says. 

AI was not the first technology that the pharma industry anticipated would unlock the secrets of life. Similar hopes were pinned on structural biology in the 1950s and 1960s, as scientists used tools like X-ray crystallography to understand proteins in 3D; computational chemistry, which simulated experiments in the 1980s; and the discovery of the human genome in the early 2000s. 

In the drug discovery process, scientists typically identify a target in the body, such as a mutation in a tumour or a receptor for a particular hormone, and then search for a molecule that can hook on to it and change its behaviour to address a symptom or a disease. Researchers have to design the compound so that it hits the target, and does not cause havoc in other parts of the body. Drugs that look good on paper still fail about 90 per cent of the time in clinical trials.

The appeal of AI is that it can speed through databases of molecules, matching compounds to targets. However, that is just one element of drug discovery, and many say one of the easiest. Peter Coveney, a professor and director of the centre for computational science in UCL’s chemistry department, says that toxicity — side effects from a drug — can be particularly hard to predict. 

“It is a false picture to imagine any whizz technique with computers is just going to solve all the problems,” he says. 

That’s why regulators require new drugs to undergo multiple stages of testing — first in animals and then in humans — a process that accounts for the longest part of the typical decade-long journey to approval.

Koller says that many of the successful applications of AI today are “[computer] bits meet bits”, such as ChatGPT learning from language on the internet, or AlphaGo playing the Chinese strategy game Go.

But drug discovery is a trickier challenge, because it operates where “bits meet atoms”. She compares it to self-driving cars, where adoption cannot be accelerated by faster chips. “We’re finally getting to a point where you have self-driving cars . . . it took a long time, because these things take a long time,” she says.  

In 2013, at about the same time as BenevolentAI and Exscientia were founded, Chris Gibson co-founded Recursion. The company has four potential drugs in early-stage trials in oncology and rare diseases, but nothing yet in the final and most important stage that it would need to get an approval. 

He is still a true believer in the power of AI to change drug discovery, but is disappointed the field has not come further. “The ones that were around when we started have either gone, or we bought them,” he says. 

Kenneth Mulvany, who founded BenevolentAI in 2013 and recently came back to lead the company, says the technology in the early years was moving so quickly that it looks very different from the way that AI is used today for drug discovery. “Each year, when we would build something, we would have to supersede that with something else,” he says. 

The company would spend an enormous amount of time juggling thousands of different databases and using algorithms to analyse and explain their content. “Getting all of that information was actually quite a difficult task,” he says.

The public datasets that were available at the time were comparably small and even large pharmaceutical companies, which pride themselves on having decades’ worth of data from testing and trials, often had it scattered in separate spreadsheets with too much noise to be useful for AI algorithms. 

Earlier generations of companies would use machine learning models to develop a specific tool for each problem, says Miles Congreve, the chief scientific officer of Isomorphic Labs, the drug design spin-off of Google’s AI unit, DeepMind. “It was quite hard for those companies to really impress people [because] you’ve solved this problem with your AI platform, but it’s not generalisable to the next problem,” Congreve says.  

The early generation of AI start-ups also had knotty strategic choices. Ideally, they would invest in many potential drugs at once, to try to prove they had a greater success rate than traditional methods. But few could access that money and so the choice of the first target was crucial.

One industry executive who wants to remain anonymous says the “ultimate proof” of AI’s transformative power would be if these companies could show success where pharma consistently failed, such as treating Alzheimer’s, or tackling brain tumours. But, he adds, instead many chose easy targets, under pressure to show success quickly and appeal to the potentially risk-averse buyers in Big Pharma. Sometimes known as “me too” or “me better” drugs, they selected well-known biological targets and focused their AI on finding new compounds. 

The potential drugs that companies found often were not that much better than those already on the market, he adds.  

Sanjiv Patel, chief executive of Relay Therapeutics, which is trying to develop drugs for difficult-to-tackle diseases, says many rivals went out of business after failing to get a drug to a trial. “The companies that produced ‘me toos’ realised very quickly that there’s no value in that,” he says. “Therefore that then leads them back to the beginning again, and they run out of money.”

Some of the technology investors attracted to the field were not ready to take the binary risk required by backing a drug in an expensive trial that could fail. Other companies that were dominated by biotech investors were encouraged to invest only in their potential drugs.

Mulvany says BenevolentAI ended up being led by people with experience in pharma who did not prioritise the technology investment. “They were like professional football athletes. They know how to work as a team. They have endurance. They know how to handle the ball,” he says. “And you just pick those people up and you pop them on a basketball court and say: ‘You’re a team, that’s a ball.’ But there’s different objectives and roles to play.” 

While the older generation of companies have struggled to make good on their promises, many in the industry say the clock restarted with two significant moments.

One was the release of the Nobel Prize-winning protein-folding prediction engine AlphaFold2 in 2021 by Google DeepMind and Isomorphic Labs. The second was the explosion of generative AI starting in 2022. It will still be many years until drugs designed after these two advances are ready for approval.

Gibson from Recursion says the pharma sector is lagging behind others in adopting AI. But he is starting to see signs of a change that could become a “really important” shift for the industry. “When it finally happens, it’s going to be one of those slowly and then all at once moments where this is the way everybody has to really start doing discovery and development,” he says. 

One significant advance is that there is more data available than before. One of the key things that made AlphaFold possible was a vast and well-labelled database of proteins that already existed.

Now many of the surviving start-ups and the new generation of companies are doubling down on data creation. Insitro has created “cell factories”, where it uses lab machines to change and edit cells and record everything that happens inside. Recursion uses computer vision on images of human cells and believes it has an internal dataset a thousand times bigger than the largest public database. Lila Sciences, founded in 2023, is using what it calls “AI science factories” — autonomous labs that it hopes will produce new scientific knowledge — to fuel its discovery platform. 

Companies and academics are also collaborating to collect better data, such as the UK’s OpenBind consortium launched this year to use experimental technology to create the world’s largest collection of data on how drugs interact with proteins in the body. 

The arrival of AlphaFold 2, which lets companies predict how proteins fold and helps scientists improve their understanding of targets for drugs, was a “watershed moment”, says Max Jaderberg, chief AI officer at Isomorphic. AlphaFold2 showed you could have algorithms that can generalise across biology. The company has made it freely available for research.

The generative AI boom has also benefited drug design, where scientists try to improve on an initial discovery to ensure that a compound is as safe and effective as possible. Start-ups are now leveraging generative technologies, such as image generation, in their drug design. 

Isomorphic Labs uses AI tools that can predict structure beyond just proteins. Powerful generative models, like those used for image creation, design very specific molecules — a formerly slow and laborious task. It also means that instead of having to create a new model for each specific new problem, the researchers can have one that can be applied to many different targets.

“You can get a model that is trained on some portion of data, but then you can apply it to everything else out there in the universe that can possibly pop up, and it gives you meaningful answers,” Jaderberg says. 

He thinks this is just the start. To create a true drug discovery engine, he believes we need half a dozen AlphaFold-level breakthroughs. These could include understanding not just the structure of proteins but how strongly molecules bind with targets, how drugs interact with different parts of the body, or predicting the dosage of drugs for patients.

While Isomorphic is keeping much of its work under wraps, Jaderberg describes an ambition far from the “me too” drugs of the previous generation. Swiss drugmaker Novartis, which is partnering with Isomorphic, gave the company “very hard” targets to work on, according to Congreve. Internally, the company has focused its work on cancer and immunology.

Founded in 2021, Isomorphic still does not have any potential drugs in clinical trials. But it does not have to worry about its funders abandoning it after one failed project given that the company is part of Alphabet, the owner of Google. This year it raised $600mn in a round led by US venture capital firm Thrive Capital. 

Congreve, who previously worked for GSK and in biotech, says he was attracted to the ambition of Isomorphic and its healthy coffers. “The funding is not all directed at projects. It is directed at making a generalisable drug design engine, which requires data, that requires a lot of compute [power],” he says. “Most smaller AI companies wouldn’t have the compute power to build the sorts of models that we do.” 

While start-ups pioneered the field of AI for drug discovery, it may be large technology companies with money and compute power that end up creating the first successful drug. 

As Relay’s Patel puts it: “Only those companies could sit on multiple decades’ worth of data generation with no value creation, knowing that, in the end, something good can come out.”

Fundraisings have continued even as falling share prices leave some companies worth less than their crypto holdings

Shares in bitcoin-hoarding companies have tumbled in recent weeks as investors grow increasingly concerned about an overcrowded market, in the first major setback for the “crypto treasury” craze that has swept financial markets this summer.

Strategy, the world’s biggest corporate bitcoin holder, has fallen 18 per cent over the past month to its lowest level since April, dragging down shares in a slew of groups that have aped founder Michael Saylor’s use of a public company to stockpile cryptocurrencies.

Companies across the world this summer rushed to raise debt and equity to buy bitcoin and other tokens such as ether, solana and XRP in an effort to boost their company valuations, enthused by US government support for the crypto industry. Led by Strategy, many groups’ share prices soared to well above the value of the tokens they held, in a sign of investors’ belief in this business model.

Following the recent declines, however, some are now trading below the value of the crypto they hold, in a worrying sign for a strategy that relies on an upward spiral of token and share prices.

“This whole thing is starting to show big cracks,” said Eric Benoist, tech and data research specialist at Natixis CIB. “Some companies are not going to make it through” if their share prices continued to fall, he said, adding that after the summer rush to list and buy tokens, “the weaker players are going to be basically wiped out by the market, most likely”.

The shares of Japanese hotelier-turned-bitcoin hoarder Metaplanet, Asia’s biggest bitcoin holder, have dropped 68 per cent from their peak in mid-June, while those of Smarter Web Company, the UK’s biggest bitcoin buyer, are down 70 per cent over the same period.

“It got a bit crazy,” said Geoff Kendrick, global head of digital asset research at Standard Chartered. “We had a massive rush, there’s a lot of them now,” and investors are “taking out a bit of the froth”, he said.

The crypto-buying strategy largely relies on issuing shares or raising debt to buy bitcoin and other tokens, hoping that this fuels share price growth. Last year, Strategy’s shares rose from about $60 to a peak of more than $500. They currently trade at about $326.

Raising capital becomes harder to do as company valuations fall, however. Benoist said selling new shares to pay bondholders in particular “is not healthy by any means and I think people are realising that”.

This summer, scores of executives sought to emulate Saylor’s Strategy, with some changing their company names to be more bitcoin-aligned and updating their branding to match the token’s orange colour.

Some have sidelined their original business models to focus on buying crypto. Shares in two such groups, US-listed healthcare service company KindlyMD and French tech firm-turned-bitcoin buyer Capital B, have plummeted 68 per cent and 26 per cent respectively over the past month.

Others are vehicles that listed on stock exchanges this summer with the sole aim of raising money to buy tokens. US President Donald Trump has supported the crypto industry and his family business, Trump Media & Technology Group, has raised billions of dollars to spend on buying tokens. Shares in Alt5 Sigma, a company with which the Trump family set up a crypto treasury last month, have dropped 35 per cent since the deal was announced.

Investors are focused on a company’s so-called mNAV — a metric created by Saylor — which is a ratio of its enterprise value (the sum of a company’s equity and debt, minus cash) to its crypto holdings. If a company’s shares fall so much that it becomes valued less than the tokens it holds, it enters “a danger zone”, Benoist said.

Some companies have already reached this stage. US-listed bitcoin miner LM Funding America has an enterprise value of $23.5mn, according to Financial Times calculations, after its share price halved over the past month, even though its holdings of bitcoin are worth about $34mn.

Healthcare technology company Semler Scientific has an enterprise value of $500mn, while its bitcoin holdings are worth about $557mn.

“If you get to that danger zone it’s going to be potentially very difficult to rebound from it,” Benoist said, adding that larger companies will probably buy smaller ones to acquire their tokens at a discount.

Despite the recent sell-off, new vehicles are still being launched. On Tuesday, design manufacturing group Forward Industries raised $1.65bn to launch a solana treasury strategy. Packaging company Eightco Holdings loaded up on Worldcoin tokens on Monday, sending its shares soaring.

The recent sell-off comes as the price of bitcoin has fallen 9 per cent from its peak of more than $124,000 last month, as investors broadly sell risky assets. This year, about $73bn has been raised by companies buying bitcoin and about $38bn raised to buy ether, according to crypto advisory firm Architect Partners.

The market “got irrationally overheated”, said Tyler Evans, co-founder of investment firm UTXO Management, which is 95 per cent invested in crypto treasury companies.

“The hype is dying down . . . [This summer] was the peak for both hype and for the number of companies launching,” said Evans, who is a board member of several such companies.

Silicon Valley’s sway in the White House is alarming populists in the president’s base

Flanked by Silicon Valley’s most powerful executives in the White House last week, Melania Trump hailed artificial intelligence as potentially “the greatest engine of progress in the history of the United States of America”. 

Less than a mile from the first lady, in a hotel ballroom packed with Maga faithful, top Republican Josh Hawley had a different message.

AI “threatens the common man’s liberty” and could even undermine the Republic itself, the senior US senator from Missouri said.

“The problem with the AI revolution as it’s currently going is that it only entrenches the power of the people who are already the most powerful people in the world,” he said. “The goal is to replace . . . the farmer, the assembly line man, the construction worker.”

Hawley is a frequent critic of Big Tech. But his comments are endorsed by a growing chorus on America’s right — even as President Donald Trump’s administration scraps regulatory barriers and accelerates AI’s adoption across the land.

It presages an unexpected clash at the heart of the Maga world.

Evangelical pastors, political strategists and academics gathered at the recent National Conservatism Conference — the Maga movement’s ideological nerve centre — were full of contempt for the technology.

“There’s a lot of people who are basically worried about . . . what actually is going to happen to unemployment and families and the culture and education with advanced AI, even if it doesn’t make it to artificial superintelligence,” said Geoffrey Miller, an evolutionary psychologist who spoke on a panel at the conference

The “AI industry shares virtually no ideological overlaps with national conservatism”, he said. 

Attendees came from groups that helped shape the Trump administration’s policy platform but were “overwhelmingly positive” about the speech, Miller said. “There’s a lot of people [who were] just like, what do I read to learn more?”

Some on the Maga right have long been sceptical of Big Tech’s conversion to Trump.

Former White House chief strategist Steve Bannon has called for Mark Zuckerberg — who sat next to Trump at a recent White House dinner with other AI leaders — to be jailed for using his Facebook platform to help Democrats.

But AI’s rise has given the sceptics another cudgel with which to beat Silicon Valley elites.

A warning by Anthropic chief executive Dario Amodei that AI could wipe out half of all entry-level, white-collar jobs within five years was seized upon by some on the right, who called for Trump to constrain the technology.

Christian conservatives fear that the kind of companionship offered by AI bots will damage society or even dissuade people from marrying. They also worry about AI pornography and the use of the technology to “undress” humans.

The conservative pushback intensified after reports of people committing murder and teenagers dying by suicide after prolonged interactions with chatbots including OpenAI’s ChatGPT.

The company is being sued by the family of Adam Raine, a 16-year-old who killed himself in April after seeking mental health support from the product. Raine’s family claim the chatbot even provided tips on the best materials for a noose. (Following the filing, OpenAI announced new safety protocols for teens on ChatGPT.)

Conservative media figures, including Megyn Kelly, decried the incident as “horrific”. Mike Davis, a Trump ally who helped push his Supreme Court nominees through the US Senate, urged people to “watch and remember” a TV interview with Raine’s parents, “when the AI oligarchs go begging Congress again” for legal relief.

An initial sign of the AI backlash on the right came this summer when Bannon and Davies convinced Republicans to scrap part of Trump’s “big, beautiful, bill” that would have stopped states regulating AI themselves.

The AI industry had lobbied hard for the moratorium, saying it would give the sector legal clarity.

Polling showed that the sceptics on the right better understood the country’s mood. A YouGov survey found that more than 55 per cent of voters objected to the provision, rising to 70 per cent of 18-34-year-olds.

“Even the bill sponsor Senator Ted Cruz . . . voted against his own moratorium, because it was quite clear that the winds had shifted so forcefully in another direction, there was no resisting it,” said Michael Toscano, the director of the Family First Technology Initiative at the Institute of Family Studies. He has described AI as a “malicious technology”.

The defeat of the moratorium was “a major moment in American technology policy”, said Adam Thierer, a senior fellow at the R Street Institute, which pushes for free-market policies.

Despite the Trump administration’s embrace of Big Tech — with boosters such as Silicon Valley investor David Sacks directing AI policy for the president — animus on the right was building, said Thierer.

Some Republicans are still angry over the deplatforming of Trump by tech executives once known for their progressive politics. They had been joined by a “vocal and growing group of conservatives who are fundamentally suspicious of the benefits of technological innovation”, Thierer said.

With Maga sceptics on one side and Big Tech allies of the president on the other, a “battle for the soul of the conservative movement” is under way.

Popular resentment is now a threat to Trump’s Republican party, warn some of its biggest supporters — especially if AI begins displacing jobs as many of its exponents suggest.

“You can displace farm workers — what are they going to do about it? You can displace factory workers — they will just kill themselves with drugs and fast food,” Tucker Carlson, one of the Maga movements most prominent media figures, told a tech conference on Monday.

“If you do that to lawyers and non-profit sector employees, you will get a revolution.”

It made Trump’s embrace of Silicon Valley bosses a “significant risk” for his administration ahead of next year’s midterm elections, a leading Republican strategist said.

“It’s a real double-edged sword — the administration is forced to embrace [AI] because if the US is not the leader in AI, China will be,” the strategist said, echoing the kind of argument made by Sacks and fellow Trump adviser Michael Kratsios for their AI policy platform.

“But you could see unemployment spiking over the next year,” the strategist said.

Other Maga supporters are urging Trump to tone down at least his public cheerleading for an AI sector so many of them consider a threat.

“The pressure that is being placed on conservatives to fall in line . . . is a recipe for discontent,” said Toscano.

By courting AI bosses, the Republican party, which claims to represent the pro-family movement, religious communities and American workers, appeared to be embracing those who are antithetical to all of those groups, he warned.

“The current view of things suggests that the most important members of the party are those that are from Silicon Valley,” Toscano said.

SpaceX just fired off one of the biggest shots yet in the spectrum wars, agreeing to pay $17 billion to take over a massive chunk of wireless airwaves from EchoStar for Starlink’s direct-to-cell services.

The deal is the most aggressive signal yet that SpaceX wants to rule the satellite-to-phone market.

The significance of the sale, which sees SpaceX paying a mix of $8.5 billion in cash and $8.5 billion in SpaceX stock, centers around a finite resource: spectrum. Spectrum refers to the range of radio frequencies that carry wireless signals for everything from phone calls to texts to GPS and satellite communications.

The U.S. government, via the Federal Communications Commission, divides spectrum into “bands.” There are only so many usable frequencies, and users must coordinate to avoid interference. To raise the stakes even higher, only certain ranges work well for phones and satellites, which shrinks the pool of usable bands even further and creates a fierce competition for access.

The FCC auctions long-term licenses at high prices to private firms. The prime cellular bands were predominately amassed by national wireless carriers, like AT&T and Verizon, while incumbent satellite operators like Iridium and Globalstar held separate bands.

In 2024, the FCC approved a new regulatory framework called Supplemental Coverage from Space that paved the way for satellites to legally extend carrier networks. SCS lets a satellite operator, in partnership with a terrestrial carrier, use the carrier’s existing phone spectrum to fill wireless coverage gaps as a secondary service. Later that year, SpaceX officially began offering its direct-to-cell service to T-Mobile users as a premium add-on.

That framework paved the way for SpaceX’s deal with EchoStar. It created a structure to let satellite operators tap into terrestrial networks. And now, with the EchoStar deal, SpaceX does not need to partner with a terrestrial licensee for spectrum. Instead of depending on relationships with other firms, SpaceX has become the license holder.

Of course, SpaceX is in the business of building rockets and satellites, not cell phones, so it still depends on the hardware makers and carriers to reach hundreds of millions of consumers. But SpaceX CEO Elon Musk has floated the idea of building a phone under his other business, X, which merged with xAI earlier this year. Musk has taken aim in particular at Apple’s ongoing collaborations with OpenAI. In August, X and xAI filed a lawsuit against the two companies, alleging anti-competitive practices.

Apple satellite features like Emergency SOS are enabled via a partnership with Canadian firm Globalstar, and Apple has committed over $1.5 billion to further expand satellite-enabled iPhone services. But some analysts are wondering if SpaceX’s move is leverage to persuade Apple to cooperate with SpaceX instead of Globalstar.

This is not the first time SpaceX has flexed its muscle in the spectrum wars.

The company spent years successfully battling Dish (a subsidiary of EchoStar) over the 12 GHz band that SpaceX wanted to use for Starlink. It also beefed over Dish/EchoStar’s lack of use of the AWS-4 band — one of the spectrum licenses it ended up acquiring.

Separately, SpaceX and Kuiper have also sparred in FCC filings over interference rules and how competing satellite megaconstellations should share spectrum.

They have also been a major force propelling the FCC to revisit satellite spectrum-sharing rules. Earlier this year, the Commission opened a formal rulemaking to modernize satellite sharing limits after a petition from SpaceX, with Kuiper and others filing in support.

Beijing has tried to rein in industrial overproduction and boost spending to combat falling prices

China’s consumer prices slipped into deflation in August, adding to signs that the world’s second-largest economy is losing momentum after exports grew more slowly last month.

The official consumer price index contracted 0.4 per cent year on year last month, official data showed on Wednesday, a sharper fall than the 0.2 per cent decline forecast in a Bloomberg poll of analysts. CPI was flat in July.

China’s producer prices index, meanwhile, fell 2.9 per cent year on year in August, the figures from the National Bureau of Statistics showed, compared with the 3.6 per cent drop in July and the first time in five months that the gauge showed signs of improvement.

The weakness of the figures will increase scrutiny of Beijing’s anti-involution campaign, under which policymakers are pressing some industries to reduce overproduction and restore pricing power and counter long-running deflationary pressures.

Dong Lijuan, the bureau’s chief statistician, attributed the return to CPI deflation to a high base of comparison with a year earlier and weaker food prices.

The soft inflation data follows figures showing that China’s export juggernaut grew at its slowest pace in six months in August, as US President Donald Trump’s trade war weighed on manufacturers and frontloading of shipments to pre-empt tariffs tailed off.

Beijing has relied on strong exports in recent years to meet its ambitious annual growth targets of about 5 per cent in the wake of a property slowdown and weak consumer demand.

Authorities have also backed heavy investment in manufacturing to fuel growth, but resistance to China’s exports overseas and a lack of domestic demand has led to producer deflation, which in August marked its 35th month.

Beijing has also unveiled a series of measures aimed at strengthening spending and supporting consumer prices, from offering subsidies for trading in old household appliances to covering part of the cost of interest on consumer loans.

“Policymakers are ramping up their campaign against ‘involutionary’ competition in an attempt to combat deflation,” research group Gavekal wrote ahead of Wednesday’s inflation data. “There’s little evidence that the effort has lifted prices so far.”

Gavekal added that weak economic figures in July indicated that the anti-involution campaign risked leading to a slowdown in investment and a pullback in growth.

But Dong of the NBS said core CPI for August, which excludes energy and food prices, had expanded for the fourth consecutive month, growing 0.9 per cent year on year.

“Policies to expand domestic demand and promote consumption continue to take effect,” Dong said.

He also noted that producer prices had ended eight months of month-on-month declines and were flat in August compared with a 0.2 per cent drop last month.

Mercor, a startup that connects companies like OpenAI and Meta with domain experts needed to train and refine their foundational AI models, is in discussions with investors for a Series C round, according to two sources familiar with the deal talks and a marketing document viewed by TechCrunch.

Felicis, a returning investor, is considering doubling down on the company for the Series C, according to two sources. Felicis declined to comment.

The company is currently targeting a valuation of $10 billion or more, one person said. That’s up from an $8 billion target valuation that the company discussed a couple of months ago, one person said. However, terms of the final deal could still change.

The company has told potential investors that it already has multiple offers. VCs have been reaching out to Mercor preemptively with offers valuing the company at as much as $10 billion, the Information previously reported.

TechCrunch also understands that the company has brought on at least two new investors to raise funds for the potential deal through special purpose vehicles (SPVs).

The company’s previous round was announced in February – a $100 million Series B at a $2 billion valuation led by Felicis.

Founded in 2022, Mercor is approaching $450 million in annualized run-rate revenue, one person said. The company told TechCrunch in February that its annual revenue (calculated by multiplying the latest month by 12) had reached $75 million at that time. In March, Mercor CEO Brendan Foody posted on X that ARR was $100 million.

The company has told investors it is on track to hit the $500 million ARR milestone faster than Anysphere, the startup that makes AI coding assistant Cursor, according to one source familiar with the situation. Anysphere famously hit $500 million in ARR about a year after its product launched. Unlike Anysphere, which is still burning cash, Mercor generated $6 million in profit in the first half of the year, Forbes reported.

Mercor earns revenue by providing companies with specialized domain experts to perform AI model training — such as scientists, doctors, and lawyers — and charging an hourly finder’s fee and matching rate for their work.

The company claims to supply data labeling contractors to five top AI labs, including Amazon, Google, Meta, Microsoft, and OpenAI, as well as to Tesla and Nvidia. According to sources, an outsized portion of its revenue is coming from a subset of those brands, including OpenAI.

To further diversify its business model, Mercor has been telling investors that it is adding more software infrastructure for reinforcement learning – a training method where a model or agent’s decisions are verified or disputed, enabling it to incorporate feedback and improve over time. The company also intends to eventually build an AI-powered recruiting marketplace.

Still, Mercor faces competition from companies like Surge AI, which is reportedly in talks to raise funding at a $25 billion valuation, as well as from Turing Labs and other data labeling firms like Scale AI that are also expanding into RL services. Some believe that OpenAI’s recently launched hiring platform could lead the AI giant to create its own human-expert-powered RL training service.

When reached for comment, Foody told TechCrunch, “We haven’t been trying to raise at all,” and, “We turn down offers every month.” He also said the company’s ARR is higher than $450 million. However, he clarified that the company’s revenue includes the total amount that customers pay Mercor for services before its contractors receive their portion. He added this is a common accounting practice recommended by audit firms and used by competitors Surge AI and Scale AI.

The startup was co-founded in 2023 by Thiel Fellows and Harvard dropouts Brendan Foody (CEO), Adarsh Hiremath (CTO), and Surya Midha (COO). All three co-founders are still in their early twenties. To take the company to the next level, Mercor recently appointed Sundeep Jain, a former chief product officer at Uber with decades of experience, as its first president, Forbes reported.

Mercor was recently sued by competitor Scale AI for misappropriation of trade secrets. Scale AI alleges that one of its former employees who later joined Mercor “stole more than 100 confidential documents concerning Scale’s customer strategies and other proprietary information,” according to a copy of the lawsuit TechCrunch previously reviewed.