>>> TradeGate Pre-Market Indications

DAX:
  • Mercedes (MBG TH) -2.6%
  • BMW (BMW TH) -3.1%
    • BMW Cuts Annual Guidance on Weak China Sales, Tariff Costs (1)
MDAX:
  • Puma (PUM TH) +2.7%
  • Aixtron (AIXA TH) -0.8%
    • Aixtron Rated New Neutral at JPMorgan; PT 13.40 euros
  • Aurubis (NDA TH) -1%
    • Aurubis Sees 2026 Pretax Operating Profit EU300M to EU400M
  • RENK Group (R3NK TH) -2%
  • Sartorius (SRT3 TH) -2.1%
    • Sartorius Cut to Hold at Berenberg; PT 225 euros
SDAX:
  • Kloeckner (KCO TH) +0.9%
  • Siltronic (WAF TH) -0.7%

>>> US After Hours Summary: PENG -13% lower on earnings; OMI +12.5% on deal to s

After Hours Summary: PENG -13% lower on earnings; OMI +12.5% on deal to sell its P&H segment; QS +6.9% higher on joint development agreement with Murata

After Hours Gainers:

Companies trading higher in after hours in reaction to earnings/guidance: None

Companies trading higher in after hours in reaction to news: OMI +12.5% (to sell its P&H segment to Platinum Equity for $375 mln), QS +6.9% (joint development agreement with Murata), IMXI +5.7% (expiration of HSR waiting period for WU acquisition of IMXI), EFX +2.6% (to sell EFX credit scores at a price more than 50% below FICO), CLPT +2.3% (results from brain tumor laser therapy study), PRIM +2.1% (names new CEO), FRPT +1.2% (CFO to step down, reaffirms FY25 outlook), TRU +0.7% (in response to EFX price cut), LEGN +0.7% (LEGN and Janssen Pharma enter into component and supply agreement), CPB +0.3% (names new CFO), DAY +0.2% (expands collaboration with MSFT; announces Dayforce AI Workspace; launches Strategic Workforce Planning; introduces new Dayforce AI Agents), LTC +0.2% (sale of two skilled nursing centers for $42 mln), BTBT +0.2% (Sept metrics), WU +0.1% (expiration of HSR waiting period for WU acquisition of IMXI), CRM +0.1% (refuses to pay hackers over data extortion attempts, according to Bloomberg)

After Hours Losers:

Companies trading lower in after hours in reaction to earnings/guidance: PENG -13% (also authorizes new $75 mln share repurchase program)

Companies trading lower in after hours in reaction to news: ESPR -18.4% (stock offering), JOBY -8.8% (files prospectus supplement for $500 mln offering), IREN -4.8% ($875 mln convertible notes offering), NMRK -4.2% (acquries RealFoundations), FICO -3.8% (in response to EFX price cut), TWNP -2.7% (stock offering by selling shareholder), CIFR -1% (reports Sept operating data), ACHC -0.8% (names new CFO), BLSH -0.5% (Sept metrics), GME -0.3% (distributes warrants), JNJ -0.2% (LEGN and Janssen Pharma enter into component and supply agreement)

Fortune : Hedge fund billionaire says 2025 is ‘so much more potentially explosiv

Hedge fund billionaire says 2025 is ‘so much more potentially explosive than 1999’ because of the way bull markets always end

Hedge fund billionaire Paul Tudor Jones, founder and CIO of Tudor Investment Corporation, has raised alarms about the state of financial markets in 2025, drawing pointed comparisons to the explosive tech-driven boom of 1999, while clarifying that today’s environment could be “so much more potentially explosive.” The reason why has to do with what the market veteran knows about how bull markets always play out.

Speaking to Andrew Ross Sorkin of The New York Times‘ DealBook on CNBC’s Squawk Box ahead of the upcoming Robin Hood Foundation investor conference, Jones described today’s investment climate as uncannily similar to the one that preceded the 2000 dot-com bust.

“It feels exactly like 1999,” he said, implying that the market was acting like the famous Prince song with the lyrics: “Party Like It’s 1999.” He urged investors to position themselves like it’s October 1999, when the Nasdaq doubled in a matter of months before collapsing, a pattern Jones sees as increasingly plausible today. The difference is that this could be even worse than the dot-com bust, he said.

Jones told Sorkin his concerns had to do with investor behavior in every bull market, specifically at the end, when the cliff is nearing and nobody is sure where it is.

The stampede
Jones underscored that in every bull market, the “greatest price appreciation is always the 12 months preceding the top,” like what happened after October 1999. For investors, he said, the challenge is timing: “If you don’t play it, you’re missing out on the juice. If you do play it, you … have to have really happy feet because there will be a really, really bad end to it,” he said, seeming to refer to the American football phrase for when a quarterback is antsy and tiptoeing around the quarterback, rather than standing their ground.

“If anything, now is so much more potentially explosive than 1999,” Tudor Jones argued, citing the backdrop of the Federal Reserve cutting interest rates. He noted the Fed is set up for several rate cuts, with monetary policy taking the economy to a real interest rate of zero, meaning incentives to invest and spend given such a low cost of capital. In terms of fiscal policy, the Congress had a budget surplus in 1999 and today it has a 6% budget deficit. “That fiscal/monetary combination is a brew that we haven’t seen since, I guess the postwar period, early ’50s, something like that,” he said. “And that was crazy times, right? Coming out of the war.”

Still, Jones insisted he wasn’t calling a bubble or predicting a crash.

“I’m not suggesting that the train is going to crash,” he clarified. “I’m suggesting that we’re in a period that’s conducive for massive price appreciation in a variety of assets.”

The ‘Boldly Go’ Era: Fiscal Recklessness and AI
Where Jones did see a bubble was in sovereign debt, labeling government bonds the “biggest bubble,” with massive global deficits staved off, for now, by an ongoing easing cycle and anticipatory flows into fixed income.

“We’re kind of boldly going where no man has ever gone before,” he said, borrowing a Star Trek reference to capture the unprecedented mix of easy money and ballooning sovereign obligations. He foresees a reckoning when the current cycle of rate cuts ends and “all that demand has been pushed forward.”


When asked about the “circularity” of financing in the AI space, with Nvidia, OpenAI, AMD, Oracle, [hotlink]Microsoft,[/hotlink] and others all having somewhat interwoven data-center deals with each other, Jones agreed. “The circularity makes me nervous, I would say.” But he added that in broad terms, the way he thinks of it is the markets are mostly concerned about inflation, and gold is a big winner (the precious metal in fact has continually set record highs in 2025, nearing $4,000 an ounce as of Monday). Jones did not mention the company Nvidia throughout his interview.

Jones also zoomed out for a minute to talk about the race for the fourth quarter.

“The race, realistically, is certainly to the end of the year, because that’s when everyone marks institutionally, and then you have to figure out what’s going to go on in next year,” he said.

These remarks echoed a separate interview given to Fortune by Bank of America Research’s senior analyst Vivek Arya.

“I think we are at that point of the year and this isn’t just specific to [2025], we have seen this in prior years right around this time where people get justifiably very nervous about what is going to be the amount of spending next year,” Arya said.

As soon as the new year starts, Arya added, “people get comfortable with spending and then it starts to get back to fundamentals again.” He added that there’s a “pecularity” about the fourth-quarter season “where people naturally get nervous.” Fortune has separately reported on Owen Lamont’s “panic season” and “harvest time” thesis, which is that markets suffer from too many traders going on vacation between August and October, a legacy that dates back to America’s agricultural roots and how markets functioned around the harvesting of certain crops.

For now, Jones told CNBC, the music is still playing. But the warning is clear: “History rhymes a lot.” For investors, 2025 could be a year defined by explosive gains—and the risk that the end of the party arrives sooner, and more violently, than most expect. He added he’s not certain whether 1999 will be entirely replaced, “but I think all the ingredients are in place, and certainly from a trading standpoint … it looks like a duck and quacks like a duck. It’s probably not a chicken, right?”

The Information : Space Startup Stoke Nears $2 Billion Valuation in New Financin

Space Startup Stoke Nears $2 Billion Valuation in New Financing

The Takeaway
  • Stoke Space in talks for funding that would double its valuation to nearly $2 billion
  • Startup develops fully reusable rockets for cost-effective launches.
  • Stoke plans first orbital rocket launch as early as next year

Stoke Space, a startup building reusable rockets that are smaller than SpaceX’s, is raising hundreds of millions of dollars in a funding round that would value it at nearly $2 billion, roughly double its last private round, two people familiar with the matter said. The new lead investor is Thomas Tull’s U.S. Innovative Technology Fund, one of the people said.

Investors consider Stoke—run by Andy Lapsa, an alum of Jeff Bezos’ rocket company, Blue Origin—a top contender among a new crop of space startups trying to crack the business of launching rockets to bring satellites to space, which SpaceX dominates. Stoke hasn’t yet launched a rocket into orbit, but plans to do so as early as next year from Cape Canaveral, Fla., Lapsa said recently.

The startup, founded in 2019 and based in Kent, Wash., is trying to reduce the costs of rocket launches by designing a rocket with engines and other parts that are fully reusable for other launches. “That’s the last big domino to knock down,” Lapsa said on a podcast in August. A Stoke spokesperson wouldn’t confirm the valuation or other details of the round.

The startup has to confront several engineering challenges to enact its vision, including creating a heat shield that will protect its Nova rocket as it reenters the Earth’s atmosphere and prevent it from burning up. The new capital, which could total as much as $500 million, should fund multiple attempts at launches, one of the people said. It would double the amount of money Stoke has raised so far, from investors including Breakthrough Energy Ventures, Glade Brook Capital Partners and Y Combinator. In January, Stoke was valued at $944 million after the investment, according to PitchBook.

Its larger contenders are also designing reusable rockets. SpaceX’s primary Falcon 9 rocket is partially reusable, with the booster that propels the rocket into space landing back on Earth for later use. SpaceX has had mixed results after 10 test launches for Starship, its rocket designed to be fully reusable that is critical to Elon Musk’s goal of colonizing Mars.

Blue Origin is also developing a reusable rocket called New Glenn. A publicly traded competitor gaining steam is Rocket Lab, which is aiming to launch a partially reusable rocket by the end of this year. Its market capitalization has nearly tripled this year to $29 billion.

Lapsa said on the podcast that he thought those companies would succeed. But, he added, “I don’t think they’re sufficient for what the commercial space economy needs and the government use case as well,” he said. “Any healthy economy needs multiple players.”

Investors have sunk billions of dollars into space-related startups, despite the setbacks or failures of some. Richard Branson’s Virgin Orbit filed for bankruptcy two years ago after the failure of a rocket launch. Privately held Sierra Space, which raised $1.4 billion from investors such as General Atlantic and Coatue Management in 2021, has faced leadership turnover and delays for its cargo-carrying space plane. And Relativity Space, backed by Fidelity Management and Mark Cuban, faced trouble raising money last year, Bloomberg reported, after its rocket failed to reach orbit after a test launch. Former Google CEO Eric Schmidt then took a controlling stake and became Relativity’s CEO and chair.

FT : Are we approaching peak prop trading?

Are we approaching peak prop trading?
Maybe! Maybe not!

Are the big trading firms that now seem to dominate markets (and whose interns earn more than the Fed chair) about to see a shift in fortunes? That’s the controversial suggestion of Alphacution’s Paul Rowady, a longtime industry analyst.

Rowady last week published an interesting report titled “Approaching Peak Proprietary”, which argued that the two primary drivers of the remarkable run enjoyed by the large trading firms will soon fade away.

The first driver has been expansion into more markets and asset classes. The big trading firms are now all active in almost every major asset class and every region, and have managed to push banks out of several of them. Here’s a killer Alphacution chart that shows the trend in US equity options (zoomable version):


For the curious, the top six banks here are (given alphabetically, rather than in order of size) Bank of America, Barclays, Credit Suisse (RIP), Goldman Sachs, JPMorgan and UBS. Alphacution doesn’t say which firms are included in “top market makers” bucket, but we can probably guess most of them.

Here’s a slightly more obtuse chart showing how dominant the big prop trading firms in particular are in cash equity trading, using positions disclosed in 13F filings as a proxy (zoomable version):


Alphacution’s report doesn’t give out the names, only anonymous ranks, but a different report from last month listed the top five as Jane Street, Susquehanna International Group, Citadel Securities, Hudson River Trading and Jump Trading, in that order.

Obviously, 13F filings only capture a moment in time, and will make some firms that typically carry more financial inventory for longer look bigger than the more pure speed merchants. For example, we’re pretty sure CitSec is still a bigger player in US equities than Jane Street, even though Jane’s disclosed holdings are more than twice as big in dollar terms.

But it’s a decent proxy for who matters, and as you can see, if you’re not top five then you’re almost inconsequential.

Things are a little bit more equal when it comes to US equity options — with the same caveat that open interest is also only a rough guide to overall trading volume shares — but not by much (zoomable version):


Rowady argues that there’s a limit to how much more market share the big prop trading firms can now profitably seize. After all, they’re now mostly competing among themselves, rather than against big old lumbering banks.

Another big driver of prop trading firm profits is that the pie itself has become bigger, even as they have increasingly eaten more of it than banks. For example, option trading volumes have exploded ever since 2020-ish, and periodic bouts of volume-boosting turmoil — eg “Liberation Day” — have been a boon to every trading desk.

But Rowady is sceptical that the growth in the velocity of trading we’ve seen lately can be sustained. The level might stay high, but it cannot grow increasingly frenzied forever.

At the same time, he notes that the prop trading industry itself is giving several oblique hints that it is near the peak:

We know that the most dominant players are near — or, at — strategy capacity for many reasons. When prop firms become active in venture capital, it’s a sign of surplus capital relative to core capacity. Another is when the smaller players consistently die off and new players are not replacing them.

The bottom line therefore is:

We don’t know when the top proprietary trading firms will hit their collective peak. They have definitely been making hay in the recent years of sunshine. Certainly, this is tied to the axes of the global listed cash and derivative product universe (or, at least, the electronically traded universe) and transaction volume. Meanwhile, the headwind to this is the concentration in just a few products — where ~40% of the US equity market has become a leveraged bet on no more than 10 stocks.

We’re inclined to agree, but only up to a point. Making markets/providing liquidity/prop trading in US equity and equity-linked markets are probably now close to zero-sum games — especially at the very fastest speeds — but they’re not quite there yet.

Moreover, the magic happens in arbitrages between all the individual stocks, equity indices and the derivatives and ETFs linked to them all. At a time of Bonk Income ETFs, it seems brave to say that equity markets are so efficient that there’s relatively less money to be made for smart traders.

We’re also sceptical about the idea that there are few inroads left to make in new asset classes and regions. Sure, there’s been lots of expansion by the big firms into areas like Asia and credit markets, but we suspect there is still plenty more potential in these areas, not least fixed income.

So, while the kind of trading revenue growth the likes of Citadel Securities, XTX and Jane Street have put up lately might not be repeated, calling the peak seems very premature.

The Information : Internal Oracle Data Show Financial Challenge of Renting Out N

Internal Oracle Data Show Financial Challenge of Renting Out Nvidia Chips

The Takeaway
  • Previously undisclosed data on Oracle AI cloud business points to industrywide challenge
  • Oracle is banking on bigger revenue to offset profit margin decreases
  • Oracle lost nearly $100 million from rentals of Blackwell chips in most recent quarter

Oracle became the best-performing megacap stock of 2025 after its executives said last month that the once-sleepy database firm will generate an astonishing $381 billion in revenue from renting out specialized cloud servers to OpenAI and other artificial intelligence developers over the next five fiscal years.

But internal documents show the fast-growing cloud business has had razor-thin gross profit margins in the past year or so, lower than what many equity analysts have estimated. That could raise questions about whether the AI cloud expansions undertaken by Oracle and its rivals will affect profitability and sustain investors’ expectations.

In the three months that ended in August, Oracle generated around $900 million from rentals of servers powered by Nvidia chips and recorded a gross profit of $125 million—equal to 14 cents for every $1 of sales, the documents show. That’s lower than the gross margins of many nontech retail businesses.

As sales from the business nearly tripled in the past year, the gross profit margin from those sales ranged between less than 10% and slightly over 20%, averaging around 16%, the documents show.

In some cases, Oracle is losing considerable sums on rentals of small quantities of both newer and older versions of Nvidia’s chips, the data show. A spokesperson for Oracle, which doesn’t publicly disclose the AI cloud server unit’s financials, did not have a comment on those figures.

The 14% gross margin figure appears to take into account the labor, power and other direct costs of running Oracle data centers, including depreciation expenses for some of the equipment. Other unspecified depreciation expenses would eat up another 7 percentage points of margin, the documents show. Equity analysts such as Gil Luria from D.A. Davidson say they tend to account for depreciation of chips and other data center equipment when calculating a cloud provider’s gross margin, as the equipment is a key cost of selling cloud services.

The internal documents offer a rare glimpse into Oracle’s dramatic transformation from a weak link of the cloud industry to an AI data center powerhouse. It also shows that being in the business of buying and renting out Nvidia graphics processing units isn’t easy.

If Oracle’s revenue from the server rental business eventually eclipses its traditional software businesses, such as databases and enterprise resource planning tools, the weak margins on server rentals mean its overall gross profit margins will drop considerably from the level they’ve been at in recent years—roughly 70%. Those margins have already sunk from around 80% a decade ago as the company has increasingly turned to selling cloud services.

The previously undisclosed Oracle figures aren’t comparable to what other AI cloud server rivals such as publicly traded CoreWeave and privately traded Lambda report to investors. Each company appears to calculate its margins differently.

All cloud firms face the same financial challenge: the high price of Nvidia’s chips. The cost of powering Nvidia server chips far exceeds that of using traditional servers, which were the backbone of the cloud computing industry before the rise of ChatGPT. AI computing also requires other specialized hardware, such as networking equipment to connect the servers.

And to win big deals with AI customers, Oracle and other cloud providers have offered hefty discounts on GPU rental prices, compared to prices they list for regular customers, according to several people with knowledge of these deals. That further diminishes the profitability of those contracts.

Several big cloud providers have said their spending on such chips has hurt their margins in recent quarters.

Building AI-focused data centers has “put pressure on Azure gross margins,” Microsoft Chief Financial Officer Amy Hood said at a Morgan Stanley event in March, in reference to the company’s cloud server rental business. But she said that pressure would lessen as Microsoft made improvements to run the servers more efficiently.

“I feel better about margins [this year] than I felt when we started in June” last year, Hood said.

Pricing ‘Aggressively’

Bigger cloud providers have been able to offset the margin hit from renting out Nvidia GPUs to some extent because most of their businesses don’t rely on that expensive hardware. Amazon Web Services reported a 33% net operating profit margin and Google Cloud reported a 17% net operating margin in the quarter that ended in June, for instance. (Neither company discloses a gross profit margin or breaks out financials for its GPU server rentals.)

Oracle is taking a more direct hit to its margins by renting out GPUs. Oracle reported around $10 billion in sales from renting out cloud servers in the fiscal year that ended in May, with around 20% of that revenue coming from GPU servers, according to the internal documents. In the most recent quarter, the percentage of GPU cloud server sales rose to 27%.

Oracle’s public disclosures imply its GPU cloud business could equal the company’s noncloud revenue as soon as 2028.

Aside from whether Oracle can build enough data centers to generate the revenue it has promised investors, another challenge is how reliant its GPU cloud business could become on a single customer.
“Oracle said, ‘OK, would you rather us not sign these deals and keep the margins higher, or would you rather us sign these deals and [see] the margins go down, but we have incremental profit?’” said Guggenheim Partners’ John DiFucci.

Oracle could still generate healthy free cash flow in the long run, even on the relatively low-margin business of renting out GPUs, if it can sustain rapid revenue growth. One way to do that could be through raising prices, something Luria reckons Oracle may attempt.

“[T]hey are pricing their offering very aggressively” right now, Luria said.

‘Not Our Specialty’

Cloud industry financial analysts largely believe AI computing is the lowest-margin business in the field, given the high cost of Nvidia GPUs. Still, Oracle’s margin on the business in recent quarters is lower than what analysts such as DiFucci and KeyBanc Capital Markets’ Jackson Ader had previously estimated.

Oracle’s margins suffer whenever it installs Nvidia’s latest GPUs in its data centers. In recent months, for instance, the GPU cloud gross margin dipped to the low teens from more than 20% due to the rollout of new Nvidia chips in facilities in Abilene, Texas, that Oracle rents out to OpenAI, the internal documents show.

Oracle could be taking an outsize hit to its margins because it doesn’t own the data centers its customers use, unlike AWS and Google Cloud, which own the majority of theirs. Instead, Oracle primarily leases its data centers from third parties, such as Crusoe.

Owning properties is “not really our specialty,” Oracle Executive Vice Chair Safra Catz said on the company’s earnings conference call last month.

Aside from whether Oracle can build enough data centers to generate the revenue it has promised investors, another challenge is how reliant its GPU cloud business could become on a single customer. Virtually all of the $317 billion worth of cloud deals it signed in the three months that ended in August came from OpenAI.

Blackwell Bleeding

Already, Oracle’s top five AI cloud customers make up roughly 80% of that business: ByteDance, Meta, xAI, OpenAI and Nvidia itself, which uses cloud-based GPUs for its own research and development. Other GPU cloud providers, including CoreWeave, Nebius Group and Lambda, face a similar risk of concentrating most of their cloud revenue in a small number of customers.

One silver lining in Oracle’s GPU business is the amount of revenue it is generating from older generations of Nvidia chips, such as the Ampere chips that came out in 2020. Those chips appear to be helping Oracle’s margins, while newer versions of Nvidia chips strain them.

That dynamic seems to contradict Nvidia CEO Jensen Huang’s recent comment that the release of newer chips essentially wipes out demand for older ones.

Oracle’s margins are also affected by how many of its servers customers are actually using—and paying for. Utilization of Oracle’s GPU cloud servers ranges between 60% and 90%, depending on the type of Nvidia chip that powers them, according to internal documents.

In the three months that ended in August, Oracle lost nearly $100 million from rentals of Nvidia’s Blackwell chips, which arrived this year. That’s partly because there is a period between when Oracle gets its data centers ready for customers and when customers start using and paying for them, the documents show. It’s not clear what causes the gap or how Oracle plans to shorten it.

The information : As Elon Musk Preps Tesla’s Optimus for Prime Time, Big Hurdles

As Elon Musk Preps Tesla’s Optimus for Prime Time, Big Hurdles Remain

The Takeaway
  • Optimus staff repeatedly talked Musk out of overambitious production goals this year
  • Tesla staff have discussed long-term plans to integrate Optimus deeply with xAI and SpaceX
  • Three executives are splitting leadership of Optimus teams since departure of engineering chief

When Elon Musk takes the stage at Tesla’s annual meeting next month, one of the centerpieces of his plan to impress shareholders will be a dancing troupe of Optimus bots, the humanoid machines that he says will eventually eclipse Tesla’s electric vehicle business, which is suffering from flagging sales growth. Behind the scenes, however, the Optimus program has been embroiled in technical problems.

Targets Musk set for Tesla to make thousands of the Optimus robots this year were abandoned by the summer, according to two people with direct knowledge of the Optimus program, a reflection of the difficulty Tesla has had with the hands for the robots.

Much is at stake. In September, Musk said Optimus would eventually account for 80% of Tesla’s value. Musk has put his own compensation on the line. At the shareholder meeting next month, shareholders will vote on whether to approve a pay package worth up to $1 trillion for Musk, which depends in part on Tesla’s success in deploying a million robots within the next 10 years. On Monday, Tesla published a video on its X account in which several versions of Optimus urged shareholders to approve Musk’s compensation package.

Musk’s ambitions for the robot cut across his empire of public and private companies. Internally, Tesla has plans to integrate the Optimus bot more deeply with Musk’s artificial intelligence startup, xAI, another person with direct knowledge of the program said. (Tesla investors will vote on a shareholder proposal that Tesla invest in xAI at the annual meeting next month, though the board of Tesla isn’t recommending a vote either way.)

Musk has also shared images of the two-legged machines walking around and building structures on Mars, where he hopes SpaceX will land a rocket in 2027. He wants Optimus to be on board a planned Mars trip, due to blast off late next year.

There’s only one issue with that plan: the current version of Optimus has been designed for indoor use on Earth, said a person with direct knowledge of the program. It’s unlikely to be able to do anything on Mars unless it wears a space suit and has a redesigned thermal management system fit for a dusty planet where the average temperature is around negative 80 degrees Fahrenheit, the person said.

Cut 2025 Targets

Tesla’s ambitions for Optimus production waned through the year. The company started 2025 with a goal of producing thousands of robots this year, two people with direct knowledge said, up from several dozen in 2024, one of the people said.

In March, Musk said at an all-hands meeting that the company would build at least 5,000 robots in 2025. But Tesla staffers dedicated to Optimus repeatedly told Musk the timelines and goals he set out to scale up production of the robot were overambitious, according to the two people with direct knowledge of the discussions.

Tesla slashed its production goal to 2,000 a few months later, one of the people said. Then, during the summer, staffers told Musk Tesla could meet the 2,000 target, but the robots wouldn’t be very useful due to issues with their hands, the most technically challenging part of their engineering, the person said. As a result, Tesla abandoned plans to produce thousands of bots this year and instead decided to further improve the robot with better hands and other design changes, the person said.

It couldn’t be learned whether Tesla has made meaningful progress since the summer, although Musk acknowledged last month on the “All-In” podcast in September that Tesla was “struggling with the final design of the hardware” and that “the hands inclusive of the forearm are a majority of the engineering difficulty of the entire robot.”

Tesla has not publicly shown the new version of the robot, which Musk has taken to calling V3. “We haven’t shown Optimus V3 yet. It is sublime,” Musk wrote on X in September.

The Optimus delays and redesign coincided with the departure of Milan Kovac, who was head of engineering for the bot until he left the company in June. Now, Konstantinos Laskaris leads Optimus hardware design, Ashok Elluswamy leads software and Lars Moravy leads manufacturing, according to one of the people with direct knowledge. While Laskaris exclusively works on Optimus, Elluswamy and Moravy also work on Tesla vehicles. All three report to Musk.

Musk said on “All-In” that he probably spends “more mental cycles” on Optimus than on any other single project. He has been intensely involved with the program, including meeting with the Optimus group at least once every week or two over the past year in Palo Alto, Calif., where he receives detailed briefings on the design and manufacturing process, two of the people with direct knowledge of the program said.

Some robotics staff at Tesla have questioned whether building a robot in humanoid form makes sense, given the complexity of designing hands—a debate that’s playing out across the industry.

A big reason Tesla is pursuing a humanoid design is that there are billions of videos of humans performing every conceivable task across the internet, which Tesla hopes to eventually use to train Optimus for every task a human can do, a person with direct knowledge said. That plan will only work if Optimus is as close to the size and shape of the average human as possible, so Tesla has designed the bot to be approximately the height of an average person, the person said.

Still, there are huge technical hurdles involved in making random videos of humans useful for training Optimus, and Tesla has so far trained the robots using videos of its own employees doing things like walking around and performing household chores.

Tesla did not respond to a request for comment.

Robot Scientists

In the near term, Tesla’s goal is to deploy as many Optimus robots as possible to perform repetitive industrial tasks inside its own factories. Optimus robots are currently doing basic tasks like sorting batteries in Tesla’s Fremont, Calif., factory, according to videos the company has shared publicly. The bots are also walking around inside the plant on security patrols, according to one of the people with direct knowledge.

Beyond the plan to send Optimus on SpaceX’s first Mars mission—which Musk wants to begin next year, though he has cautioned there may be delays—staff have also discussed other futuristic ways to use the robot through integration with another one of his companies, xAI.

Optimus can already use the Grok chatbot made by Musk’s xAI to speak, a feature Musk has touted on X. Eventually, Tesla envisions using more-advanced versions of Grok and Optimus to replace white-collar workers at the company, including scientists, according to a person with direct knowledge. Tesla staff have discussed having Optimus one day perform hands-on experiments in laboratories using a more advanced and specialized version of Grok as its brain, the person said.

In the meantime, Tesla is facing mounting competition from other companies including Meta Platforms and Amazon, both of which are pushing into humanoid robotics. There are also several other startups, including Figure AI and 1X, that are raising billions of dollars to build humanoids for home and industrial uses.

At The Information’s AI Agenda Live summit in New York last week, 1X CEO Bernt Børnich was asked what advantage his company has over rivals like Tesla and Meta Platforms. “The simple answer is a working product,” he said.