On Friday afternoon, right as this interview commenced, a news notification popped up on my screen: the Trump administration was cutting connections with Anthropic, the AI firm based in San Francisco that was established in 2021 by Dario Amodei. Shortly thereafter, Defense Secretary Pete Hegseth invoked national security legislation to blacklist the company from engaging with the Pentagon after Amodei declined to permit Anthropic’s technology to be utilized for widespread surveillance of American citizens or for autonomous armed drones capable of selecting and eliminating targets without human oversight.

It was an astonishing turn of events. Anthropic risks forfeiting a contract valued at up to $200 million and may be prohibited from collaborating with other defense contractors after President Trump shared a post on Truth Social instructing every federal agency to “immediately halt all usage of Anthropic technology.” (Anthropic has since stated it will contest the Pentagon’s actions in court.)

Max Tegmark has dedicated much of the past decade to warning that the competition to develop increasingly potent AI systems is advancing quicker than the world’s capacity to manage them. The MIT physicist founded the Future of Life Institute in 2014 and in 2023 played a key role in drafting an open letter — ultimately endorsed by over 33,000 individuals, including Elon Musk — calling for a halt in the development of advanced AI.

His perspective on the Anthropic situation is unflinching: the company, like its competitors, has contributed to its own dilemma. Tegmark argues that the issue starts not with the Pentagon but with a decision made years ago — a collective choice within the industry to resist regulation. Anthropic, OpenAI, Google DeepMind, and others have long asserted their commitment to self-regulation. This week, Anthropic even abandoned the cornerstone of its own safety assurance — the promise not to release increasingly powerful AI systems until it was assured they wouldn’t pose a risk.

Now, without established regulations, there isn’t much safeguarding these entities, Tegmark notes. Here’s more from that discussion, condensed for brevity and clarity. You can listen to the entire dialogue next week on TechCrunch’s StrictlyVC Download podcast.

When you learned of the news regarding Anthropic, what was your initial response?

The path to disaster is lined with good intentions. It’s fascinating to reflect back a decade, when people were filled with optimism about how we would harness artificial intelligence to cure cancer, enhance prosperity in America, and strengthen the nation. And now we find ourselves with the U.S. government angered at this company for not wanting AI to be employed for domestic mass surveillance of Americans, and also opposing the creation of killer robots that can independently — without any human intervention — determine whom to kill.

Anthropic has built its entire identity around being a safety-first AI company, yet it was engaged with defense and intelligence organizations [tracing back to at least 2024]. Do you find that at all contradictory?

It is indeed contradictory. If I may offer a somewhat cynical perspective — yes, Anthropic has been quite adept at marketing itself as safety-oriented. However, when one scrutinizes the reality rather than the assertions, it becomes apparent that Anthropic, OpenAI, Google DeepMind, and xAI have all spoken extensively about their dedication to safety. None have advocated for binding safety regulations like those found in other industries. All four of these companies have now failed to uphold their own promises. First, we witnessed Google with its grand slogan, ‘Don’t be evil.’ Then they abandoned that. Following that, they discarded another broader commitment that essentially stated they would not inflict harm with AI. They moved away from this in order to sell AI for surveillance and military applications. OpenAI recently removed the term safety from its mission statement. xAI disbanded their entire safety team. And now Anthropic, earlier this week, renounced its most critical safety commitment — the pledge not to deploy advanced AI systems until it was assured they wouldn’t cause damage.

How did entities that made such significant safety pledges end up in this predicament?

All of these companies, particularly OpenAI and Google DeepMind, but to a degree Anthropic as well, have consistently lobbied against AI regulation, asserting, ‘Just trust us, we’ll self-regulate.’ And they’ve been effective in their lobbying efforts. Consequently, we currently have less oversight over AI systems in America than we do over sandwiches. For instance, if you wish to open a sandwich shop and a health inspector discovers 15 rats in the kitchen, they won’t permit you to sell any sandwiches until rectified. Conversely, if you claim, ‘Don’t worry, I won’t sell sandwiches; I’m going to offer AI girlfriends for 11-year-olds, which have been associated with suicides previously, and then I’m going to launch something called superintelligence that could potentially overthrow the U.S. government, but I feel good about mine’ — the inspector has to respond, ‘Alright, proceed, just avoid selling sandwiches.’

There are food safety regulations but no AI regulations.

And I hold all of these companies collectively accountable for this. Because if they had taken seriously all the promises they made previously about being safe and virtuous, and united to approach the government, asking, ‘Please transform our voluntary commitments into U.S. law that mandates even our most negligent competitors’ — this situation might have been avoided. Instead, we’re facing a comprehensive regulatory void. History has shown us what results from complete corporate immunity: thalidomide, tobacco firms marketing cigarettes to children, asbestos leading to lung cancer. It’s somewhat ironic that their own steadfastness against establishing laws delineating acceptable and unacceptable practices regarding AI is now rebounding against them.

There currently exists no law prohibiting the creation of AI to kill Americans, so the government can simply demand it. Had the companies previously advocated for, ‘We desire this law,’ they wouldn’t find themselves in this predicament. They effectively shot themselves in the foot.

The counter-argument from the companies is always centered around the competition with China — if American firms don’t pursue certain paths, Beijing will. Does that argument hold true?

Let’s dissect that. The most prevalent talking point from the lobbyists representing AI companies — who are now better funded and more numerous than the combined lobbyists from the fossil fuel, pharmaceutical, and military-industrial sectors — is that whenever any regulatory proposals are made, they say, ‘But China.’ So let’s examine that. China is currently working on banning AI girlfriends outright. Not merely age limits — they’re contemplating a ban on all anthropomorphic AI. Why? Not to appease America, but because they believe this is detrimental to Chinese youth and makes China vulnerable. Clearly, it’s also weakening American youth.

When advocates state we must compete to develop superintelligence to gain the upper hand against China — while we lack clarity on how to control superintelligence, leading to the potential of humanity losing command of the Earth to alien machines — it’s apparent that the Chinese Communist Party favors control. Who in their right mind believes Xi Jinping will allow a Chinese AI entity to develop something that could topple the Chinese regime? Absolutely no chance. This scenario also poses a substantial threat to the American government — if it were to be overthrown by the first American enterprise to achieve superintelligence. This constitutes a national security risk.

That’s a compelling perspective — framing superintelligence as a national security risk instead of an asset. Do you perceive this viewpoint gaining traction in Washington?

I believe that if individuals within the national security sector hear Dario Amodei portray his vision — he delivered a notable speech where he mentioned that we’ll soon have a nation of geniuses housed in a data center — they might start pondering: ‘Wait, did Dario just mention the word nation? Perhaps I should include that nation of geniuses in a data center in the same threat assessment I’m tracking, as that sounds perilous for the U.S. government.’ And I expect that in the near future, a sufficient number of individuals in the U.S. national security sector will recognize that uncontrollable superintelligence is a threat, not merely a tool. This situation is entirely analogous to the Cold War. There was a quest for supremacy — economically and militarily — against the Soviet Union. We, as Americans, won that round without engaging in the subsequent conflict, which would determine who could create the most nuclear craters in the opposing superpower. People realized that such an approach was suicidal. No one prevails. The same reasoning is relevant in this context.

What implications does this hold for the overall pace of AI advancement? And how near do you think we are to the systems you’re discussing?

Six years ago, nearly every AI expert I was acquainted with predicted we were decades away from developing AI capable of mastering language and knowledge at human levels — perhaps 2040, perhaps 2050. They were all mistaken, as we currently possess that capability. AI has advanced swiftly from high school proficiency to college, PhD level, and even to university professor standards in certain domains. Last year, AI secured the gold medal at the International Mathematics Olympiad, which represents one of the most complex human tasks. I co-authored a paper with Yoshua Bengio, Dan Hendrycks, and other leading AI researchers just a few months back, providing a rigorous definition of AGI. By this definition, GPT-4 was at 27% of the way there. GPT-5 was at 57% of the way. Therefore, we aren’t there yet, but the leap from 27% to 57% so rapidly indicates that it may not be long before we reach that point.

When I lectured to my students yesterday at MIT, I informed them that even if it takes four years, when they graduate they might find job opportunities severely limited. It’s certainly time to begin preparations.

With Anthropic now on the blacklist, I’m intrigued to observe what unfolds — will the other AI giants support it and declare, ‘We won’t pursue this either?’ Or will a company like xAI step forward and state, ‘Anthropic didn’t want that contract; we’re on board’? [Editor’s note: Hours after the interview, OpenAI announced its own agreement with the Pentagon.]

Last night, Sam Altman stated he stands with Anthropic and shares the same ethical boundaries. I respect him for his bravery in voicing that stance. Google, at the time this interview commenced, had yet to issue any response. If they remain silent, it would be profoundly embarrassing for them as a corporation, and much of their workforce is likely to feel similarly. We also haven’t received any word from xAI yet. It will be fascinating to witness how this unfolds. Essentially, this moment requires everyone to reveal their true colors.

Is there a scenario in which the outcome could be favorable?

Yes, and this is why I find myself oddly optimistic. There’s such a clear alternative available. If we begin to treat AI companies like we do any other enterprises — eliminating corporate amnesty — they would necessary undertake something akin to a clinical trial prior to releasing such potent products, demonstrating to independent experts their capability to manage it. In turn, we could usher in a golden age marked by the benefits of AI, devoid of existential dread. Although that’s not the trajectory we’re currently pursuing, it remains a possibility.

This week, Netflix shocked the entertainment industry by opting not to increase its offer for Warner Bros. Discovery, paving the way for Paramount Skydance to acquire the Hollywood studio.

At that moment, Netflix co-CEOs Ted Sarandos and Greg Peters expressed their commitment to financial prudence. Now, reporting from Bloomberg provides additional insights into why Netflix officials stepped away from a bidding contest that appeared to be in their favor last December. 

One reason is that the streaming behemoth’s investors seemed highly doubtful about the acquisition being a wise move — Netflix’s stock plummeted 30% after the deal was announced, while the news of its retreat caused Netflix shares to rise nearly 14%.

Additionally, Netflix’s resolve regarding the deal reportedly weakened after Paramount delivered a higher bid and showed readiness to continue with multiple rounds of bidding.

By the time Sarandos met with officials from the Trump administration on Thursday, he might have already made the choice to back down. Indeed, since President Donald Trump had previously cautioned him against overpaying, Sarandos reportedly remarked, “I took your advice.”

Meanwhile, employees at Warner Bros. are now concerned about significant layoffs at the studio and heightened conservative political influence on CNN.

The entertainment and streaming sector has just experienced one of its most significant megadeals to date, leaving industry analysts astonished. This not only marks a historic financial milestone, but it is also forecasted to shake up Hollywood and the entire media landscape as we currently perceive it. 

Following years of Warner Bros. Discovery grappling with an overwhelming burden of billions in debt, coupled with dwindling cable viewership and intense rivalry from streaming services, the firm has been weighing substantial strategic alternatives, such as divesting its entertainment properties to a competitor.

Various major stakeholders recognized the opportunity to acquire the media behemoth, and in December, Netflix announced its intent to purchase WBD’s studios and streaming operations for $82.7 billion.

However, in a shocking last-minute development this month, it appears the Paramount, under David Ellison’s leadership, may emerge victorious in this bidding conflict, proposing $111 billion for all of Warner Bros. Discovery’s assets, encompassing its studios, HBO, streaming services, games, and television networks like CNN and HGTV. Paramount was recently acquired by Ellison, with substantial backing from his father, Larry Ellison, chairman of Oracle, the sixth-richest individual in the world and a significant donor to Trump.

Paramount’s proposal is still pending formal endorsement from WBD’s board of directors, and any potential arrangement may also encounter regulatory scrutiny.

Let’s examine what is unfolding, what implications are at play, and what the next steps could entail. 

What has transpired thus far?

​This situation originated back in October, when Warner Bros. Discovery (WBD) disclosed its consideration of a possible sale due to unsolicited interest from various key industry players.

​The bidding process quickly escalated in competitiveness, with Paramount and Comcast identifying themselves as serious bidders, with Paramount initially seen as leading the charge. 

Nonetheless, WBD’s board ultimately recognized Netflix’s offer as the most appealing, with Netflix proposing $82.7 billion for only Warner’s film, television, and streaming properties.

This initiation sparked the bidding war. Paramount contended that its offer of roughly $108 billion for all of Warner’s assets was superior to Netflix’s more limited focus on studios and streaming. To enhance its proposal, Netflix revised its offer in January to a full cash deal at $27.75 per share of Warner Bros. Discovery, providing further assurance to investors and facilitating the path for the deal to advance.

​Paramount continued its pursuit of acquiring WBD, yet the Warner board consistently turned down its proposals, expressing worries about Paramount’s substantial debt and the associated risks of its offer, including concerns regarding the extensive array of investors funding Paramount’s bid, such as sovereign wealth funds from Saudi Arabia, Qatar, and Abu Dhabi. The board highlighted that Paramount’s offer would burden the merged entity with $87 billion in debt, a risk they were not prepared to assume at that time.

In January, Paramount initiated a lawsuit to obtain more clarity on the Netflix deal. A month later, it attempted to enhance its offer by proposing a $0.25 per share “ticking fee” to WBD shareholders for each quarter the agreement remains unclosed by December 31, 2026. It additionally indicated it would cover the $2.8 billion breakup fee if Warner were to withdraw from its arrangement with Netflix.

Subsequently, in a final maneuver to secure a transaction, Paramount upped its bid to $31 per share in February. This compelled the WBD board to extend talks with Paramount regarding a possible deal, viewing it as a more advantageous proposal. Netflix chose not to elevate its offer and exited the discussions.

“The deal we brokered would have generated shareholder value with a discernible path to regulatory approval,” said Netflix co-CEOs Ted Sarandos and Greg Peters in a statement on Feb. 26. “However, we have consistently maintained discipline, and given the price necessary to align with Paramount Skydance’s latest offer, the transaction is no longer financially viable, thus we are opting out of matching the Paramount Skydance proposal.”

In addition to the billions in debt Paramount currently holds, the company is also poised to take on the roughly $33 billion in debt that Warner Bros. Discovery has under the proposed agreement. This acquisition will be supported by a $54 billion debt commitment from Bank of America Merrill Lynch, Citi, and Apollo Global Management, along with $45.7 billion in equity from Larry Ellison.

Regulatory challenges and additional concerns

Alongside the substantial debt assumption that represents a considerable financial strain, Paramount is also confronted with various other challenges related to its negotiation with WBD that could affect the success of the proposed transaction. 

For starters, Ellison has cautioned about considerable job cuts expected shortly. Critics have raised alarm over the likelihood of job reductions and diminished wages.

Ellison also carries a controversial reputation in the sector, and his stewardship of CBS News has been perceived as aligned with support for Donald Trump’s administration, an association strengthened by the substantial contributions from his father, Larry Ellison. Under Ellison’s management at Paramount, critical reporting against the administration has been marginalized or scrutinized closely by Ellison or his appointed CBS News head, the conservative provocateur Bari Weiss.

This has sparked concerns among staff at Warner-owned CNN. Trump has personally demanded concessions from news outlets that are critical of him, including a $16 million settlement from CBS, as a prerequisite for his FCC approving the Ellison takeover of Paramount. Prior to Netflix withdrawing from negotiations, Trump pressured the company to remove former Biden White House official Susan Rice from its board, publicly expressing intentions to bring CNN to heel under new ownership.

Regulatory scrutiny also presents another obstacle. Such a major merger has drawn attention from lawmakers.

For instance, California Attorney General Rob Bonta stated on February 26 that “these two Hollywood powerhouses have not overcome regulatory examination — the California Department of Justice is currently investigating, and we plan to be rigorous in our review.”

A day before Netflix withdrew its bid, it was disclosed that a coalition of 11 state attorneys general urged the U.S. Department of Justice (DOJ) to analyze the merger amid concerns it could suppress competition and inflate subscription rates. This follows months of U.S. senators Elizabeth Warren, Bernie Sanders, and Richard Blumenthal expressing their apprehensions to the Justice Department’s Antitrust Division, warning that such a massive merger may yield serious ramifications for consumers and the industry as a whole. The senators argue that the merger could grant the new media conglomerate excessive market power, enabling it to raise consumer prices and hinder competition.

Nevertheless, Larry Ellison, Oracle’s chairman and a notable Trump donor, has close connections to the Trump administration. His acquisition of Paramount last year proceeded swiftly after acquiescing to compliance.

When is the deal anticipated to finalize?

The agreement is not yet concluded.

Initially, a deal with Netflix was expected to lead to a stockholder vote around April, with a conclusion anticipated within 12 to 18 months following that vote. However, the shift to the Paramount deal will likely reset the timeline for approvals. Furthermore, regulatory endorsements are still pending, and scrutiny could influence the ultimate result. 

Stay tuned…

Claude, the chatbot created by Anthropic, appears to have gained from the heightened interest in the firm’s contentious discussions with the Pentagon.

As initially disclosed by CNBC, as of Saturday afternoon, Claude holds the second position among free applications in the US App Store by Apple — with OpenAI’s ChatGPT in first place and Google Gemini in third.

Based on analytics from SensorTower, Claude was just on the cusp of the top 100 at the end of January and has spent the majority of February within the top 20. Its position has surged recently, moving from sixth on Wednesday to fourth on Thursday and then to second on Saturday (today).

Following Anthropic’s attempts to secure measures that would prevent the Department of Defense from utilizing its AI technology for mass domestic surveillance or entirely autonomous weaponry, President Donald Trump instructed federal agencies to cease usage of all Anthropic products, with Secretary of Defense Pete Hegseth categorizing the company as a supply-chain threat.

In response, OpenAI announced its own deal with the Pentagon, which CEO Sam Altman asserted incorporates protections concerning domestic surveillance and autonomous weaponry.

A significant amount of computational capacity is necessary to operate an AI product — and while the tech sector quickly seeks to harness the capabilities of AI models, there is a simultaneous endeavor to construct the infrastructure that will support them. In a recent earnings call, Nvidia CEO Jensen Huang projected that by the decade’s end, investments in AI infrastructure will reach between $3 trillion and $4 trillion, primarily from AI enterprises. This endeavor is exerting tremendous pressure on power grids and pushing the industry’s construction capacity to its edge.

Below, we have compiled all known details regarding the largest AI infrastructure initiatives, including substantial investments from Meta, Oracle, Microsoft, Google, and OpenAI. We will continue to update this information as the boom persists and the figures rise further.

Microsoft’s 2019 investment in OpenAI

This is arguably the agreement that initiated the entire modern AI surge: In 2019, Microsoft invested $1 billion in an exciting non-profit known as OpenAI, primarily recognized for its connection with Elon Musk. Importantly, this arrangement designated Microsoft as OpenAI’s exclusive cloud service provider — and as the demands for model training intensified, a larger portion of Microsoft’s investment began to be delivered as Azure cloud credits instead of cash.

It proved beneficial for both parties: Microsoft was able to report increased Azure sales, while OpenAI received additional funding for its major expenditure. In the ensuing years, Microsoft would elevate its investment to nearly $14 billion — a strategic move likely to yield significant returns once OpenAI transitions to a profit-driven entity.

The relationship between the two firms has unraveled more recently. Last year, OpenAI declared it would no longer exclusively utilize Microsoft’s cloud, offering the company a right of first refusal for future infrastructure needs while exploring other options if Azure couldn’t accommodate them. Microsoft has also started pursuing alternative foundational models for its AI products, aiming for greater autonomy from the AI behemoth.

OpenAI’s success with Microsoft has prompted a trend among AI services to partner with specific cloud providers. Anthropic has secured $8 billion in funding from Amazon, modifying the hardware at the kernel level to better support AI training. Google Cloud has also partnered with smaller AI firms like Lovable and Windsurf as “primary computing partners,” although these agreements didn’t include direct investments. OpenAI has additionally returned for more support, garnering a $100 billion investment from Nvidia in September, enabling it to acquire a greater number of the company’s GPUs.

The rise of Oracle

On June 30, 2025, Oracle disclosed in an SEC filing that it had entered into a $30 billion cloud services agreement with an unnamed partner; this exceeds the company’s total cloud revenues for the preceding fiscal year. OpenAI was ultimately disclosed as the partner, positioning Oracle alongside Google as one of OpenAI’s series of hosting partners post-Microsoft. Unsurprisingly, the company’s stock surged.

A few months later, another significant announcement was made. On September 10, Oracle unveiled a five-year, $300 billion arrangement for computing power, set to commence in 2027. Oracle’s stock ascended even further, briefly positioning founder Larry Ellison as the wealthiest individual globally. The immense scale of this deal is astonishing: OpenAI does not possess $300 billion to allocate, suggesting significant growth expectations for both companies, coupled with a degree of optimism.

However, even before any funds are utilized, the arrangement has solidified Oracle’s place as a primary AI infrastructure provider — and a formidable financial entity.

Nvidia’s investment spree

As AI laboratories rush to create infrastructure, their primary source for GPUs has been one company: Nvidia. This trading has resulted in Nvidia accumulating substantial capital — which it has reinvested into the industry in progressively unconventional ways. In September 2025, Nvidia acquired a 4% stake in competitor Intel for $5 billion — yet even more surprising have been the agreements with its own clients. A week subsequent to the Intel transaction, Nvidia announced a $100 billion investment in OpenAI, compensated with GPUs that would be allocated to OpenAI’s ongoing data center initiatives. Nvidia has since revealed a comparable deal with Elon Musk’s xAI, while OpenAI initiated a separate GPU-for-equity arrangement with AMD.

If that seems circular, it’s because it is. Nvidia’s GPUs are coveted due to their scarcity — and by exchanging them directly into an ever-expanding data center initiative, Nvidia is ensuring they remain that way. The same can be said for OpenAI’s privately held stock, which becomes increasingly valuable as it is unavailable through public markets. Currently, both OpenAI and Nvidia are thriving and there’s little anxiety — but should the momentum begin to wane, this set-up will face significantly more examination.

Building tomorrow’s hyperscale data centers

For organizations like Meta, which already possess substantial legacy infrastructure, the situation is more intricate — though equally costly. Meta CEO Mark Zuckerberg has stated that the firm intends to invest $600 billion in U.S. infrastructure by the end of 2028.

In the first half of 2025, the company allocated $30 billion more than the previous year, primarily driven by its expanding AI goals. Part of that expenditure was directed towards high-value cloud contracts, such as a recent $10 billion agreement with Google Cloud, but even more resources are being funneled into two enormous new data centers.

A new 2,250-acre location in Louisiana, named Hyperion, is anticipated to incur an estimated $10 billion in construction costs and is expected to provide around 5 gigawatts of computing power. Significantly, the site includes a partnership with a local nuclear power facility to accommodate the augmented energy demand. A smaller facility in Ohio, referred to as Prometheus, is slated to become operational in 2026, fueled by natural gas. 

That type of expansion comes with tangible environmental repercussions. Elon Musk’s xAI constructed its own hybrid data center and energy-generation plant in South Memphis, Tennessee. The facility has swiftly become one of the county’s largest emitters of smog-generating chemicals, due to a series of natural gas turbines that experts assert violate the Clean Air Act.

The Stargate moonshot

Just two days following his second inauguration last January, President Trump announced a collaborative initiative involving SoftBank, OpenAI, and Oracle, aiming to invest $500 billion in the development of AI infrastructure within the United States. Dubbed “Stargate” after the 1994 film, the project arrived with immense anticipation, with Trump labeling it “the largest AI infrastructure endeavor in history.” OpenAI’s Sam Altman seemed to concur, stating, “I believe this will be the most significant project of this era.” 

In basic terms, the proposal involved SoftBank providing the financing, while Oracle managed the construction with guidance from OpenAI. Trump oversaw the entire operation, pledging to eliminate any regulatory obstacles that might delay the project’s progress. Yet, doubts surfaced from the onset, including from Elon Musk, Altman’s business adversary, who alleged the initiative lacked the necessary funding.

As the initial excitement has subsided, the project has experienced a decline in momentum. In August, Bloomberg reported that the collaborators were struggling to reach agreement. Nevertheless, the initiative has continued with the construction of eight data centers in Abilene, Texas, with the final building expected to be completed by the end of 2026.

The capex crunch

“Capital expenditures” usually refers to a company’s investment in physical assets, a rather dull metric. However, as tech companies prepared to announce their capex forecasts for 2026, the surge in data center spending transformed the figures into something much more engaging — and significantly larger.

Amazon led the pack in capex, projecting $200 billion in spending for 2026 (up from $131 billion in 2025), while Google closely followed with estimates ranging from $175 billion to $185 billion (up from $91 billion in 2025). Meta estimated expenditures of $115 billion to $135 billion (up from $71 billion the previous year), although this figure is somewhat misleading as many data center projects have been omitted from their financial reports. Overall, hyperscaler firms are planning to invest nearly $700 billion in data center initiatives in 2026 alone.

This influx of capital has caused some concern among investors. Most companies, however, remain unfazed, emphasizing that AI infrastructure is crucial for their future. This has created a peculiar dynamic. As expected, tech leaders are more optimistic about AI than their Wall Street counterparts — and the larger the investments tech companies make, the more apprehensive their bankers become. When added with the considerable debts many firms are incurring to finance these expansions, it’s not uncommon to hear CFOs throughout Silicon Valley grinding their teeth.

Although this has yet to slow AI expenditures, it may soon — unless, of course, hyperscalers can demonstrate that these investments yield substantial returns.

This article was initially published on September 22.

On Friday evening, OpenAI CEO Sam Altman revealed that the company has finalized a deal permitting the Department of Defense to implement its AI models within the department’s secure network.

This development follows a prominent confrontation between the DoD — referred to as the Department of War during the Trump era — and OpenAI’s competitor Anthropic. The Pentagon urged AI companies, including Anthropic, to allow their models to be utilized for “all lawful purposes,” while Anthropic aimed to establish a boundary concerning mass surveillance and entirely autonomous weapons.

In an extensive statement issued on Thursday, Anthropic CEO Dario Amodei asserted that the organization “never objected to specific military operations nor attempted to restrict usage of our technology in an ad hoc way,” but maintained that “in a limited number of instances, we believe AI can compromise, rather than uphold, democratic ideals.”

This week, over 60 OpenAI employees alongside 300 employees from Google endorsed an open letter urging their companies to back Anthropic’s stance.

Following the unsuccessful negotiations between Anthropic and the Pentagon, President Donald Trump condemned the “Leftwing nut jobs at Anthropic” in a social media update that also instructed federal agencies to cease utilizing the company’s products after a six-month wind-down period.

In another post, Secretary of Defense Pete Hegseth alleged that Anthropic was attempting to “usurp veto authority over the operational choices of the United States military.” Hegseth further stated that he is categorizing Anthropic as a supply-chain risk: “Effective immediately, no contractor, supplier, or partner engaging with the United States military may perform any commercial operations with Anthropic.”

On Friday, Anthropic stated it had “not yet received direct communication from the Department of War or the White House regarding the status of our negotiations,” but emphasized it would “contest any supply chain risk designation legally.”

Unexpectedly, Altman mentioned in a post on X that OpenAI’s new defense contract includes safeguards that address the same concerns that became contentious for Anthropic.

“Two of our key safety principles are bans on domestic mass surveillance and human accountability for the use of force, including for autonomous weapon systems,” Altman stated. “The DoW concurs with these principles, manifests them in law and policy, and we incorporated them into our agreement.”

Altman noted that OpenAI “will establish technical safeguards to ensure our models function correctly, which the DoW also required,” and it will assign engineers to the Pentagon “to assist with our models and to ensure their safety.”

“We are urging the DoW to extend these same conditions to all AI companies, which we believe should be acceptable for everyone,” Altman added. “We have conveyed our strong wish to see a reduction in legal and governmental confrontations, moving towards reasonable solutions.”

Fortune’s Sharon Goldman reports that Altman informed OpenAI staff during an all-hands meeting that the government would permit the company to develop its own “safety stack” to avert misuse and that “if the model declines to perform a task, then the government would not compel OpenAI to execute that task.”

Altman’s post was published just before it was revealed that the U.S. and Israeli governments have commenced bombing Iran, with Trump advocating for the ousting of the Iranian administration.

Xiaomi has introduced a variety of devices ahead of the Mobile World Congress (MWC) in Barcelona, which includes a flagship smartphone aimed at photography enthusiasts, a clone of AirTag, the Xiaomi Watch 5 smartwatch, and a remarkably slim power bank.

The company, headquartered in China, has teamed up with Leica to co-brand the Xiaomi 17 Ultra smartphone. This collaboration involves utilizing Leica lenses and crafting filters inspired by the renowned German camera manufacturer.

Equipped with a 50-megapixel main sensor featuring an F/1.67 aperture and a 1-inch sensor, the standout feature is the 200-megapixel telephoto camera offering a variable focal length equivalent to 75mm-100mm. This enables optical zoom capabilities ranging from 3.2x to 4.3x. Additionally, there’s a 50 MP ultrawide camera boasting an f/2.2 aperture.

Moreover, the device comes with a robust 6,000 mAh battery (with the Chinese variant featuring a larger 6,800mAh capacity). It can be charged at 90W via USB PD-PPS and also supports Xiaomi’s Hypercharge wireless technology at 50W.

The smartphone boasts a 6.9-inch Xiaomi HyperRGB OLED display safeguarded by Xiaomi’s Shield Glass 3.0. The device is powered by Qualcomm’s Snapdragon 8 Elite Gen 5 processor, which is also featured in the newly released Galaxy S26 series.

Additionally, a special Leica edition phone commemorating 100 years of the camera brand will be available. This edition features a resilient aluminum-alloy body with a nickel-anodized finish, and it includes a Leica theme on the software side.

This device features a rotating ring that simulates zooming akin to a traditional camera. The special edition is also equipped with a “Leica Essential mode,” incorporating filters that evoke the photographic style of the Leica M9 and Leica M3.

The Xiaomi 17 was introduced with an increased 6,330 mAh battery, capable of charging at 100W utilizing the company’s HyperCharge technology.

Furthermore, two photography accessories for the Xiaomi 17 Ultra are being unveiled. The Xiaomi 17 Ultra Photography Kit is a Bluetooth-enabled attachment featuring a two-stage shutter button alongside a video recording button. The Xiaomi 17 Ultra Photography Kit Pro aims to replicate a traditional camera appearance with a leather finish, a video recording button, a detachable shutter button, and zoom functionality. This accessory connects via USB-C and is equipped with a 2,000 mAh battery to operate. Users can also utilize a new fastshot mode with this attachment.

With this launch, the company is making the devices accessible in the EU and the UK. The Xiaomi 17 will have a starting price of €999, while the Xiaomi 17 Ultra begins at €1,499. The Leica edition, featuring 16GB RAM and 1TB storage, will be priced at €1,999. The Xiaomi 17 Ultra Photography Kit costs €99.99, and the Xiaomi 17 Ultra Photography Kit Pro is set at €199.99.

In addition to smartphones, the company unveiled numerous other gadgets, including a new scooter. Xiaomi’s Electric Scooter 6 Ultra offers a peak power of 1200W and a range of 75km. It comes equipped with 12-inch all-terrain tires and both front and rear disc brakes. A three-inch TFT display provides essential metrics such as speed and range. The scooter starts at €329.99 in five versions, with the highest-priced model at €799.99.

Additionally, the company introduced a new Xiaomi tag, a device akin to AirTag, compatible with both Apple Find My and Google Android Find Hub. Weighing merely 10 grams, it features a button cell battery lasting over a year. Users can play a sound remotely to locate the tag or note the time when it is attached. The pricing for this tag is set at €14.9 for a single unit and €49.99 for a bundle of four.

Moreover, the company presented a sleek power bank measuring only 6mm in thickness. Weighing 98 grams, it has a battery capacity of 5,000 mAh. It supports charging devices at 22.5W via wired connection and at 15W wirelessly. The magnetic power bank can adhere to compatible phones like iPhones and charge them wirelessly. For pricing, it’s set at €59.99 in black and silver, while the orange variant costs €64.99.

Xiaomi also announced its latest smartwatch, the Xiaomi Watch 5, which features a 930mAh battery capable of lasting up to six days. This smartwatch comes with a 1.54-inch AMOLED display and supports gestures for dismissing calls or alarms. It can generate a health report in 60 seconds by measuring various metrics such as heart rate, blood oxygen, stress levels, sleep duration, sleep heart rate, and sleep SpO₂. The price for this watch is set at €299.99.

Additionally, the company presented the €69.9 Redmi Buds 8 Pro earbuds, which feature active noise cancellation and offer up to 33 hours of battery life.

China’s humanoid robots captured worldwide interest with kung fu stunts during the nation’s aired Spring Festival Gala, while Chinese smartphone manufacturer Honor is poised to introduce its inaugural humanoid robot at MWC in Spain. 

Robotics has been identified as a key focus in the country’s “Made in China 2025” initiative, originally aimed at factory automation but now shifting towards humanoid robots. Significant advancements in multimodal AI are boosting what is referred to as embodied AI — autonomous machines functioning in real-world settings — a move that officials claim could alleviate labor shortages and enhance productivity. 

At this nascent phase of humanoid robot evolution, Chinese firms are surpassing their U.S. competitors in both speed and production volume, stated Selina Xu, an AI policy lead in Eric Schmidt’s office.

“China boasts a more established hardware supply chain — largely developed through the electric vehicle industry, including sensors and batteries — and the world’s most powerful manufacturing infrastructure, enabling companies to innovate significantly faster than Western counterparts,” Xu informed TechCrunch. 

Consequently, Chinese robots are not only more cost-effective but companies can also launch new models at a quicker pace, Xu observed, noting that top Chinese company Unitree delivered approximately 36 times more units last year than American competitors Figure and Tesla.  

In the previous year, global humanoid robot shipments reached merely 13,317 units, as per a Forbes report released last month. This is a minuscule figure for an industry projected to nearly double every year, reaching 2.6 million units by 2035. (However, caution is necessary when interpreting these figures. The report indicates uncertainty regarding how many units represent actual commercial sales compared to demo units or pilot trials, highlighting the industry’s early-stage status.) 

The leading humanoid robot manufacturers by 2025 shipment volumes were China’s Agibot and Unitree, succeeded by UBTech, Leju Robotics, Engine AI, and Fourier Intelligence, highlighting Beijing’s preliminary supremacy in the field.  

Recently, the most significant shift has transitioned from “demo-driven enthusiasm” to “operations-driven implementation,” Yuli Zhao, chief strategy officer at Galbot, shared with TechCrunch. Galbot’s humanoid robot, the G1, was showcased at this year’s Spring Festival Gala, China’s yearly state-sponsored lunar New Year’s Eve broadcast, alongside robots from Unitree Robotics, Noetix, and MagicLab. 

“An increasing number of customers are inquiring: Can the robot operate consistently in real-life scenarios and genuinely assist people? This practical demand is reinforced in China, where policy and industrial strategies promote automation upgrades, and the manufacturing environment enables rapid iteration,” Zhao remarked.

While boosted financing for humanoid startups “has certainly accelerated” progress, “the most sustainable adoption occurs when you can demonstrate consistent and reproducible value in production or service settings, rather than just a one-time exhibition,” Zhao added.

Nevertheless, investments are crucial, and Chinese robotics manufacturers are securing them. Last year, Unitree was valued at approximately $3 billion after concluding its Series C, with aspirations of reaching as high as $7 billion in a potential IPO. Concurrently, Galbot has attracted over $300 million in new investments, reportedly elevating its valuation to $3 billion, marking one of the largest funding rounds in China’s humanoid robotics sector to date. 

U.S. firms are also advancing beyond impressive demonstrations to emphasize real-world implementations. Furthermore, they are pursuing their ambitious targets. U.S. startup Foundation intends to develop 50,000 humanoid robots by the end of 2027. 

However, China is already aiming for a combination of budget-friendly mass-market models and premium applications, quickly proliferating humanoids across industrial, consumer, and rehabilitation domains, as highlighted in a December TrendForce report.

Obstacles to China’s dominance

Regarding AI systems and integrated software, the true standing of Chinese humanoid companies remains ambiguous. The industry is fundamentally relying on vision-language-action frameworks and “world models,” yet both technologies are still in their infancy. Nvidia currently dominates this area with its comprehensive humanoid software suite, according to Xu, leading most humanoid startups in China to depend on Nvidia’s Orin chips. Nonetheless, local chip manufacturers are working on developing native alternatives, she noted. 

Yet, humanoid robotics developers are still tackling fundamental issues. The challenge lies in enabling robot foundational models to anticipate the “next physical state” they will encounter in unpredictable settings, similar to how large language models project the next word. Unlike LLMs, however, humanoid robotics firms cannot merely harvest internet data for training, Xu explained. Thus, most are relying on simulated environments to create synthetic data, although gathering real-world data is crucial.

“Due to the data scarcity issue, humanoids are still quite a distance from true autonomy. The hardware is presently ahead of the software — the robot bodies can now manage substantially more dexterity than in the past (although they face reliability issues, as seen when some robots malfunctioned during humanoid marathons), but the cognitive functions remain in the early stages,” the analyst stated. 

Safety is also a major obstacle for humanoid robots. A single high-profile incident could lead to public backlash, and China is likely considering how to swiftly deploy the technology without proceeding too hastily. As the sector matures, stronger regulations are anticipated.

Given the data limitations, Zhao posits that the initial demand for humanoids will likely grow first in relatively controlled work environments.

“Early momentum is expected in industrial manufacturing, warehouse logistics, and retail, where tasks are repetitive, hours are lengthy, and processes are well-defined — generating genuine demand and optimal conditions for humanoid robots to deliver large-scale value,” he stated. 

Additional APAC competitors 

Humanoid robot advancement is not confined to a two-nation competition. Japan’s robotics sector — encompassing startups to semiconductor powerhouses — aims for humanoid mass production by 2027. Long a trailblazer with initiatives like Honda’s Asimo, Murata Manufacturing’s Murata Boy, and SoftBank Robotics’ Pepper, Japan relies on precision and advanced control. A unique area for this country: Humanoid robots are increasingly utilized in eldercare.  

Coral Capital CEO James Riney, who invests in technology firms in Japan, is confident that Tokyo will continue to prosper in the humanoid robotics field. “Three factors are likely to propel the adoption of robotics in Japan. One is the labor shortage and the intent to rely less on mass immigration. The second is the prevalent cultural perception of robots as allies — more Doraemon than Terminator. The third is that Japan already excels in many aspects of the robotics supply chain.”

Hyundai Motor’s Boston Dynamics division unveiled a new Atlas humanoid scheduled for factory use by 2028, aiming to manufacture up to 30,000 units yearly in the U.S. as part of its AI-driven robotics strategy.  

For China, however, government initiatives, industrial strategies, labor deficiencies, and private investments are converging to accelerate the country’s humanoid robotics initiative. 

 “China’s leadership can be best understood as a speed-to-scale advantage,” Zhao remarked. “The ecosystem here compresses the entire cycle — research and development, supply chain, manufacturing, integration, and customer deployment — into a very tight loop. This allows humanoid firms to transition from prototype to real-world deployment more swiftly, learn from actual operations, and iterate at a pace that is challenging to replicate elsewhere.”