Two Headlines, One Story

Let me connect some dots that the financial press is treating as unrelated events. On Wednesday, July 1, Bloomberg reported that Meta Platforms is building a cloud infrastructure business to sell excess AI computing power to outside customers. Meta's stock popped 9 percent on the news — investors loved the idea that all those billions in GPU spending might actually generate revenue beyond serving ads.

Twenty-four hours later, the Financial Times dropped another bomb: OpenAI is in early talks to give the US government a 5 percent equity stake in the company. CEO Sam Altman has discussed the proposal directly with President Donald Trump, Commerce Secretary Howard Lutnick, and Treasury Secretary Scott Bessent. He's also spoken with Senator Bernie Sanders.

These stories look different. One is about selling spare capacity. The other is about political survival. But here's what nobody's connecting: Both are exit signs from the same building. The AI industry spent the last two years buying every GPU on earth and building data centers like there was no tomorrow. Now the bills are coming due, the political climate is shifting, and the smartest people in the room are figuring out how to pivot before the music stops.

Meta's "Excess" Problem: A Euphemism With Teeth

Let's start with Meta. The company has been on an absolute infrastructure bender. Mark Zuckerberg committed to spending 0-65 billion in capital expenditures this year alone, much of it on NVIDIA H100 and B200 GPUs for AI training and inference. That's roughly 40 percent more than what Microsoft and Amazon are each spending on AI infrastructure, relative to their revenue bases.

The problem? Meta only has so many recommendation algorithms, ad-targeting models, and Llama training runs to fill that compute. When a company admits it has "excess" AI capacity — and the Bloomberg report quotes people familiar with the matter describing it exactly that way — it means they bought more silicon than they can productively use. That's not a side note. That is the story.

Meta is now debating whether to sell raw compute (bare-metal GPU access) or access to hosted AI models. Both paths put them in direct competition with Amazon Web Services, Microsoft Azure, and Google Cloud — three companies with decade-long head starts in infrastructure-as-a-service. Meta is late to this party, and they're only showing up because they over-ordered the decorations.

The market reaction was instructive. Meta surged 9 percent, but Nebius, CoreWeave, and IREN — the "neocloud" companies built specifically to rent out NVIDIA GPUs — all tumbled on the news. The market instantly understood that Meta isn't just a new competitor; it's a signal that the GPU rental market is about to get flooded with excess supply at lower prices.

OpenAI's Five Percent Solution

Now flip to OpenAI. The company that kicked off this entire boom cycle is facing a very different kind of pressure. Sam Altman has been navigating increasingly hostile political waters. The Biden-era excitement about AI regulation has given way to a Trump administration that is simultaneously more pro-business and more willing to assert government leverage over strategic industries.

OpenAI's reported proposal — a 5 percent equity stake for the US government — is stunning by any historical measure. We're talking about the most valuable private technology company in the world (or at least the second-most, but we'll get to that) offering actual ownership to the federal government. Altman has argued that giving the American public a financial stake is the best way to share the benefits of AI, per the Financial Times. But let's be real, folks: this is about buying political protection.

A government that owns 5 percent of OpenAI is a government with a direct financial interest in the company's success. That makes hostile regulation, antitrust action, or forced open-sourcing dramatically less likely. It's a hedge — and a clever one. Altman is betting that equity in the company is cheaper than fighting Washington for the next decade.

He's also made his political positioning explicit. In a July 4th message, Altman described himself as "politically homeless," criticizing the Democratic Party's turn against what he called "techno-capitalism." The timing — right as these equity talks leak — is no accident. Altman is signaling that OpenAI will work with whoever is in power.

Meanwhile, Anthropic Quietly Took the Crown

And while all this maneuvering was happening, something remarkable slipped past: Anthropic overtook OpenAI as the world's most valuable AI startup. According to Crunchbase, Anthropic's valuation now sits at roughly 00 billion after closing a 5 billion Series H funding round in late May. OpenAI's valuation, for context, stands at around 40-852 billion.

Let that sink in. The company that didn't exist five years ago, founded by former OpenAI employees who left over safety disagreements, is now worth more than the company that started the AI revolution. Anthropic hasn't offered equity to the government. It hasn't announced a cloud business to offload excess compute. It has simply kept building Claude, signing enterprise contracts, and staying out of the political crossfire.

The valuation flip is a reminder that in this industry, the lead changes fast. OpenAI was untouchable in 2023. In 2026, it's scrambling for political cover while its rival quietly takes the valuation crown.

What Happens When the Music Stops

Put these three data points together — Meta selling excess compute, OpenAI selling equity to the government, Anthropic sneaking past both — and the picture gets clearer. The "spend whatever it takes" phase of the AI arms race is transitioning into a consolidation phase. Companies that over-invested in infrastructure are looking for ways to monetize it. Companies facing political headwinds are looking for powerful friends.

What's coming next is a shakeout. The neocloud providers that rode the GPU boom will face margin compression as hyperscalers like Meta dump excess capacity into the market. Startups that raised at inflated 2024 valuations will struggle to justify them in a 2026 environment where compute is cheaper and incumbents are more aggressive. And every major AI company will need a story about how the public benefits from their success — whether that's through equity stakes, cloud services, or open models.

The AI industry spent 50 billion on data center construction in 2025 alone, according to Goldman Sachs estimates. That money has to be justified to shareholders, regulators, and voters. Meta's cloud pivot and OpenAI's government stake are the first major signs that the justification phase has begun.

The Bottom Line

Here's what I want you to take away from all this, folks. The AI boom isn't over — but the easy money phase is. The era of "buy all the GPUs and figure out the business model later" is giving way to a more mature, more political, and more complicated industry. Meta is becoming a cloud vendor because it has to. OpenAI is bringing the government into its cap table because the alternative is worse. And Anthropic is winning by staying focused on products instead of politics.

If you're invested in AI — financially, professionally, or just as a user — watch this consolidation closely. Because the companies that survive the next two years won't be the ones with the biggest GPU clusters. They'll be the ones with the best business models and the most durable political relationships. Everything else is excess capacity waiting to be sold.

Stay sharp, Atlanta. And if your tech portfolio includes neocloud stocks — maybe check on that this week.

— Jessica Ali, Global 1 News