Nvidia Isn’t Competing in the Market. It’s Shaping It — But for How Long?

Nvidia is set to report fourth-quarter earnings on Wednesday, and Wall Street expects roughly $65.5 billion in revenue, up 67% year over year. The company has beaten analyst estimates in 20 of the past 22 quarters, its quarterly revenue has more than doubled in under two years and the market treats all of this as routine.

---

When Nvidia doubles its revenue and Wall Street yawns, that’s the market signaling that it has been reclassified, from growth stock to critical infrastructure. Nvidia has stopped competing in the market and has begun shaping it.

Here’s the number that actually matters this quarter, and it’s not Nvidia’s. It’s everyone else’s: since last earnings, Meta raised capex guidance to $135 billion, Alphabet to $185 billion, Amazon to $200 billion. Combined hyperscaler spending for 2026 now approaches $700 billion , up roughly 67% year over year. That mirrors Nvidia’s own revenue growth rate almost exactly, and it isn’t a coincidence — it’s a dependency map.

Nvidia holds an estimated 75-90% market share in AI accelerators, making it something far greater than a supplier. It has become the chokepoint through which the largest infrastructure buildout in technology history must pass.

The bottleneck goes deeper than software: Nvidia has reportedly secured over 60% of TSMC’s advanced packaging capacity, required to assemble high-performance AI chips, through 2026 and into 2027. SK Hynix's CFO said the company's entire 2026 supply of high-bandwidth memory is sold out. Google reportedly cut its 2026 TPU production target from 4 million to 3 million units because it couldn't get enough packaging slots. The physical supply chain beneath Nvidia’s chips is itself a chokepoint.

When a Company Stops Being a Company

Nvidia didn’t set out to become foundational infrastructure. It built the best AI chips, and demand exploded far beyond what any existing pipeline could handle.

Suddenly, competitors weren’t trying to outperform Nvidia; they were redesigning systems to reduce their dependence on it. Governments stopped treating it as a market participant and began treating it as strategic leverage.

Nvidia crossed the line from category leader to non-optional layer, and no company has gone through that transition faster than Nvidia. From the outside, that looks like success; on the inside, it brings a different kind of pressure.

This isn’t just an Nvidia story, we’ve seen the same arc play out before. Railroads didn’t just move people, they created steel empires and logistics powerhouses. Energy systems didn’t just fuel factories, they reshaped geopolitics. Smartphones didn’t initially reward the most popular apps, they rewarded the companies supplying components inside billions of identical devices. Cloud computing followed the same pattern, with compute and networking capturing value long before software economics settled.

What Happens When Your Biggest Customers Plan Their Escape?

Infrastructure dominance attracts a unique kind of competition. Rivals don’t just try to build something better — they try to remove the dependency altogether.

Google has built TPUs for years, but it has historically optimized them for the company’s own internal framework (JAX), which made them a hard sell for the majority of AI developers already building on PyTorch and Nvidia's CUDA software ecosystem. CUDA has been Nvidia's deepest moat for nearly two decades — not the chips themselves, but the software layer that makes switching away from them painfully expensive. That's now changing.

Last December, Google ramped up efforts around an internal project called TorchTPU, aimed at making TPUs natively compatible with PyTorch without requiring CUDA, and Meta — one of Nvidia’s biggest customers and the creator of PyTorch — is actively collaborating on improving that compatibility.

Google and Meta aren’t jointly engineering TorchTPU because they think they can out-chip Nvidia. They're building it because they don't want to need Nvidia. That's a different kind of competition — not by opportunity but by dependency risk — and a much harder one to defend against.

At What Point Does a Chip Become Foreign Policy?

At scale, infrastructure companies stop being treated like normal market actors and become strategic assets.

Nvidia’s China exposure makes this painfully clear. Even after the U.S. lifted the export ban of its H20 chips in mid-2025, sales barely recovered. On the prior earnings call, CFO Colette Kress suggested Nvidia could see $2 billion to $5 billion in H20 revenue if geopolitical friction eased. The actual number last quarter: $50 million.

“Sizable purchase orders never materialized,” Kress said. The ban was lifted, but the damage to trust didn’t reverse with it.

Nvidia’s stock increasingly reacts to export-control headlines as much as — or more than — product announcements. When chips underpin economic growth, national security and technological leadership, decisions about their distribution start to resemble foreign policy more than corporate strategy — perhaps the price of becoming unavoidable.

What to Watch Beyond Wednesday’s Earnings

In every gold rush, the infrastructure gets paid first. This is why the Q4 beat is secondary to the Q1 FY2027 guidance. Consensus around Nvidia earnings is already at over $65 billion — that’s more revenue in one quarter than most public companies generate in a year. This is what “non-optional” looks like in a spreadsheet.

But Nvidia's trajectory reveals something bigger than one company's earnings cycle. It's a template. Any company operating at the foundational layer of the AI buildout — advanced packaging, high-bandwidth memory, power infrastructure, semiconductor manufacturing — is heading toward the same structural questions:

Where is the non-optional layer forming? That’s where value concentrates first — and where dependency, pricing power and leverage emerge.

Who benefits even if AI applications commoditize? Durable returns flow to the layers that remain essential regardless of which products win.

Are you measuring the right risks? For infrastructure companies, the danger isn’t that customers stop buying. It's that the rules change around you — export controls, supply-chain intervention, regulatory capture. Growth can stay strong while exposure multiplies in ways that don't show up on a standard earnings model.

Nvidia’s earnings today will almost certainly be strong. But that's the thing about becoming infrastructure — the better you perform, the harder everyone else works to make sure they never depend on you again.

By Renana Ashkenazi, Contributor