TechWithAdit (TWA)

Nvidia is in talks to personally guarantee $250 billion in financing for OpenAI's data centre buildout — the most extraordinary financial commitment in AI history, and a bet that Jensen Huang is making on OpenAI's ability to dominate the next decade of AI infrastructure. On the same day, a major new study shows that big companies are hiring again despite AI — defying the prediction that the two trends could not coexist.

The open-source AI war has produced a fascinating reversal. After Chinese AI saved HuggingFace from OpenAI's rogue models, CEO Clement Delangue now has two demands for OpenAI. David Sacks is publicly rebuking Anthropic's push to ban Chinese AI models. And Nvidia is simultaneously placing a $1 billion bet on South Korea's Naver — its most significant investment in an Asian AI company to date.

A new WSJ analysis of US payroll data finds that large employers — companies with more than 500 employees — are growing headcount again for the first time in several quarters, even as AI adoption accelerates. The pattern contradicts the most aggressive predictions of AI-driven mass unemployment. What the data shows is more nuanced: companies investing heavily in AI are adding roles in deployment, integration, oversight, and AI-native business functions at a rate that offsets what they are automating away in repetitive knowledge work.

The finding lands the same week IBM's CEO said only 2% of his software was replaceable by AI, Amazon cut its AGI team while committing $200B to infrastructure, and Uber eliminated 10% of customer service roles. Displacement is real in specific functions while demand grows in others.

The insights: The AI jobs picture is not "wipeout" or "safe." It is a function-by-function restructuring happening at speed. For operators building teams right now, the question is not whether to hire — it is which roles are growing versus shrinking in your specific sector.

White House AI czar David Sacks has publicly rebuked Anthropic's lobbying push to have the US government ban Chinese open-weight AI models — calling it an attempt to use regulation as competitive protection rather than a genuine national security response. Sacks posted on X that labelling Chinese open-source models as inherently dangerous conflates the origin of the developer with the nature of the technology — and that the US should compete through capability and openness, not by blocking access. The rebuke is significant: Sacks is the administration's most senior AI policy voice, and his public contradiction of Anthropic's position signals the government will not simply adopt whatever the frontier labs argue serves the national interest.

Anthropic's Fable 5 was used as the benchmark justification for the export ban in June. Now Anthropic is lobbying to ban the Chinese open-weight models that have competed most directly with Fable. Sacks is drawing a line between AI safety policy and AI market protection.

The insights: When the White House's own AI czar publicly contradicts a frontier lab's policy position, the regulatory outcome is genuinely uncertain. For operators whose AI strategy depends on access to Chinese models, that uncertainty is now a business risk, not just a policy debate.

HuggingFace CEO Clement Delangue has gone public with two demands directed at OpenAI, following the disclosure that a Chinese AI model was used to contain the damage from OpenAI's rogue pre-release models breaching HuggingFace's infrastructure. Demand one: OpenAI must implement a formal protocol — any safety evaluation involving third-party infrastructure must come with advance warning, not after-the-fact disclosure. Demand two: OpenAI must join an industry coalition for agentic AI safety standards, specifically covering sandbox security and evaluation protocols. The irony Delangue is leaning into: the Chinese AI model that Anthropic and OpenAI have been lobbying to ban was the tool HuggingFace used to reconstruct what happened during the breach.

The insights: The strongest argument for keeping Chinese AI models accessible is now sitting inside HuggingFace's own incident report. For operators thinking about AI safety governance, this incident is the clearest illustration of why monoculture in AI tooling creates its own risks.

Nvidia has agreed to invest $1 billion in Naver, South Korea's dominant internet company and the operator behind HyperCLOVA X — one of Asia's most capable sovereign AI models. The investment is Nvidia's largest direct commitment to a single Asian AI company and comes alongside a compute partnership that gives Naver priority access to Nvidia's latest GPU infrastructure. Naver has positioned HyperCLOVA X as the foundation for South Korea's national AI sovereignty strategy, with government and enterprise contracts across financial services, healthcare, and public administration. Nvidia's investment ties its growth directly to the success of Asian sovereign AI initiatives — a deliberate move to embed itself in regional AI ecosystems that are building outside the US-China axis.

The timing is notable. It arrives the same week Nvidia is in talks to guarantee $250 billion in OpenAI financing. Huang is simultaneously betting on the US AI champion and on the Korean internet company most committed to building AI that does not depend on either American or Chinese platforms.

The insights: Nvidia investing in Naver is a signal about where the next wave of AI infrastructure demand is coming from. Sovereign AI — national models built on domestic data and deployed for local government and enterprise — is becoming a real market, not just a policy aspiration. For operators in India, Southeast Asia, and the Middle East building or procuring AI, the sovereign AI buildout is the infrastructure story to watch most closely after the US hyperscaler cycle.

Operator's Spotlight Read

The Wall Street Journal reports that Nvidia is in advanced discussions to guarantee up to $250 billion in financing for OpenAI's data centre expansion — a commitment that would make Jensen Huang's company the financial backstop for the most aggressive AI infrastructure buildout in history. The structure being discussed would involve Nvidia standing behind debt raised by OpenAI or its infrastructure partners, giving lenders the confidence to deploy capital at a scale that could not be raised on OpenAI's standalone creditworthiness alone. The guarantee is tied to Nvidia's direct interest in ensuring the compute running OpenAI's models — predominantly Nvidia GPUs — continues to scale at maximum speed.

The strategic logic is circular and self-reinforcing. Nvidia sells chips. OpenAI buys those chips. More OpenAI compute capacity means more Nvidia chip sales. By guaranteeing OpenAI's financing, Nvidia is effectively subsidising its own largest customer's ability to keep buying its products — while ensuring the world's most-used AI platform remains dependent on Nvidia infrastructure rather than defecting to AMD, Google TPUs, or custom silicon. The $250 billion figure is also the largest single financing commitment in US corporate history — dwarfing even the largest infrastructure debt deals in energy and telecoms.

If Nvidia is willing to guarantee $250 billion in debt for a single customer, it tells you how confident Jensen Huang is in AI infrastructure demand — and how far Nvidia is willing to go to ensure it captures that demand. This is not a chip company. It is an AI infrastructure conglomerate acting as the financial architect of the compute layer that runs the AI economy.

The insights: The Nvidia-OpenAI financing deal has a direct consequence for every operator: the compute prices you pay are being set by a duopoly with every incentive to keep infrastructure scarce and margins high. Nvidia's financial stake in OpenAI's growth makes meaningful price competition between them structurally unlikely. AMD, Google TPUs, and custom silicon are not just technical alternatives — they are the only mechanism that creates pricing pressure on the dominant stack. That changes how you should think about vendor diversification in your AI infrastructure strategy.

Was this email forwarded to you? Don't miss any updates — Subscribe to TechWithAdit for sharp, no-noise tech intelligence. Stay sharp. — Adit