Sam Altman Walks Back the Jobs Apocalypse, Anthropic Closes In on a Trillion-Dollar Valuation

TLDR: Sam Altman now says AI-driven mass job displacement is not happening at the pace he once feared, fewer entry-level white-collar roles have disappeared than predicted. Meanwhile Anthropic is on the verge of closing a $30B funding round at a $900B valuation, and humanoid robots just signed their first real commercial warehouse deal.

Summary:

Let's start with the headline that will make a lot of people feel conflicted. Sam Altman, the person who spent years warning that AI could devastate the labor market, is now saying the jobs apocalypse isn't really materializing. He specifically called out entry-level white-collar work as a category where losses have been smaller than anticipated. That is worth sitting with for a moment. This is the same person whose company builds the product that employers are pointing to when they freeze junior hiring. The claim that displacement is less severe than feared does not match what a lot of early-career knowledge workers are actually experiencing when they try to find their first job. Altman may be right on the macro numbers while still being wrong about the lived reality for people entering the workforce right now.

On the money side of the industry, Anthropic is reportedly closing a $30 billion funding round that would push its valuation to $900 billion. That would make it more valuable than OpenAI, which has been the dominant name in the space. A $900 billion valuation for a company that has not yet turned a profit demands some skepticism. These numbers are being driven by investor competition for a small number of assets in a perceived winner-take-most market, not by current fundamentals. That said, Anthropic has made real technical progress, and their enterprise positioning has been deliberate and careful compared to competitors.

Microsoft shipped something genuinely interesting this week. MAI-Image-2.5 is their first in-house image generation model, and it debuted at number three on the LMArena text-to-image leaderboard. Microsoft has largely been a distributor of other people's AI, integrating OpenAI models across their product surface. Building their own image model signals they want more vertical control, particularly as the relationship with OpenAI gets more complicated over time. Debuting at third place is a real result, not a participation trophy.

Anthropic also announced Claude Managed Agents, which can run inside self-hosted sandboxes and connect to private MCP servers within enterprise security boundaries. This is a direct response to the very real concern that enterprises have about sending sensitive workloads to third-party cloud infrastructure. Running agents that stay inside your own security perimeter, with access to internal tools via MCP, solves a problem that has been blocking AI adoption in regulated industries. The timing makes sense given where enterprise conversations are headed.

Finally, Figure AI signed a commercial deal with Catalyst Brands, the parent company of JCPenney, to deploy humanoid robots at a distribution center in Reno. This is the first commercial humanoid deployment at this scale in a US warehouse. Humanoid robots have been perpetually "18 months away" from real deployment for years. An actual signed contract with a named customer and a specific location is a different category of event than a demo video.

Key takeaways:

Sam Altman now downplays AI job displacement, but his framing may not reflect what entry-level workers are actually experiencing in the job market
Anthropic is approaching a $900B valuation, which would make it more valuable than OpenAI, though the number reflects investor competition more than current revenue
Microsoft shipped its first in-house image model and landed at number three on a major benchmark, suggesting a gradual move toward less dependency on OpenAI
Claude Managed Agents running in self-hosted sandboxes connected to private MCP servers addresses a genuine enterprise blocker around data security
Figure AI signed the first real commercial humanoid robot contract in a US warehouse, moving the category from demo to deployment

Why do I care: The Claude Managed Agents announcement is the most directly relevant item here for anyone building enterprise software. The ability to run agents inside your own infrastructure and connect them to internal tools via MCP is exactly the architecture that large customers require before they will let AI touch anything sensitive. If you are building or advising on enterprise AI adoption, this is the pattern worth understanding now. The Microsoft image model is worth watching as a signal, not because MAI-Image-2.5 changes anything today, but because it suggests Microsoft is building capability to reduce OpenAI dependency over time, which has downstream implications for the whole ecosystem.

☕🤖 Sam Altman: The AI "Jobs Apocalypse" Isn't Coming