Nvidia's Biggest Customers Are Quietly Building Their Own Chips to Escape It

Here is a quiet revolution hiding behind the AI boom. Nvidia sells the chips that power almost all of AI, and business is booming. But its four biggest customers, Google, Amazon, Microsoft, and Meta, are spending tens of billions of dollars to build their own chips so they can stop depending on Nvidia. And for the first time, in 2026, these custom chips are growing faster than Nvidia's. It is one of the most important power struggles in technology, and almost nobody outside the industry is watching it. Here is what is happening and why it matters.

One grew revenue 50x after half his team quit over the strategy. One brought in 50K signups in a single day with no paid budget. One generated 100M+ views from a stunt that took 50 hours to conceive. One asked every prospect to demo the product themselves instead of demoing it for them.

None of them followed the safe playbook. They treated GTM like an experiment, moved before they had proof, and made bets most founders would never get approved.

HubSpot for Startups documented all 6 stories in the free Bold Bets Playbook. The risks they took, why it was risky, and what it returned.

Read the free Bold Bets Playbook

For the last few years, the story of AI has been the story of Nvidia. Its chips power nearly all of the AI you use, and demand has been so high that its sales are, in the company's own words, off the charts. But behind that dominance, its biggest customers have been plotting their independence.

Here is the shift. Custom AI chips are outpacing Nvidia GPU shipment growth in 2026 for the first time, with custom chip shipments projected to grow 44.6% against 16.1% for the merchant GPUs that Nvidia sells. The companies driving this are the giants you would expect. Google's Ironwood TPU, Amazon's Trainium series, Microsoft's Maia 200, and Meta's MTIA accelerators are all custom chips that reduce their builders' dependence on Nvidia hardware and optimize for the specific work each company runs at scale. Den of GeekDen of Geek

The reason these companies are willing to spend billions building their own chips comes down to simple math, and one real-world example makes it vivid. Midjourney, the AI image platform, reported cutting monthly compute costs from approximately $2.1 million to $700,000 after moving its work from Nvidia GPUs to Google's custom chips, a 65% reduction. When you can cut your computing bill by nearly two-thirds, and you are running billions of AI tasks a day, building your own chips stops being a science project and becomes an obvious financial decision. Den of Geek

CNBC's explainer comparing the top AI chips: https://www.cnbc.com/2025/11/21/nvidia-gpus-google-tpus-aws-trainium-comparing-the-top-ai-chips.html

Tech Times on custom chips outpacing Nvidia: https://www.techtimes.com/articles/317225/20260526/custom-ai-chips-outpace-nvidia-gpu-growth-2026-asic-shipments-set-triple-gpu-rate.htm

Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.

Try Wispr Flow free

This is not a small side project. The scale of what these companies have built is genuinely staggering.

Take Amazon. Amazon built Project Rainier, an internal AI computing cluster constructed entirely from its own Trainium chips, containing 500,000 custom processors in a single system. For context, OpenAI used approximately 25,000 Nvidia GPUs to train GPT-4. Project Rainier is 20 times that, all on Amazon's own custom silicon. That is a company replacing the most valuable product in tech, at massive scale, with hardware it designed itself. euronews

Google is even further along. Google has moved into mass deployment, reporting that over 75% of its Gemini model computations are now handled by its internal chip fleet. That is remarkable. The Gemini model that, as we have covered, is about to power both Android phones and Apple's new Siri runs mostly on chips Google built, not on Nvidia's. Microsoft is doing the same for its products. Microsoft's custom Maia 200 chip currently serves AI models for OpenAI and powers Microsoft 365 Copilot workloads. GM AuthorityFremontleaf

The trend is so strong that analysts are forecasting a dramatic shift. Experts predict that by the end of the decade, over 60% of all AI compute will run on non-Nvidia hardware, a total reversal of the market dynamics we saw just three years ago. GM Authority

The Silicon Sovereignty analysis on hyperscalers breaking Nvidia's grip: https://markets.financialcontent.com/stocks/article/tokenring-2026-1-5-the-silicon-sovereignty-era-hyperscalers-break-nvidias-grip-with-3nm-custom-ai-chips

Here is the part of the story that reveals where the real power, and the real money, is quietly flowing. The hyperscalers design these chips, but they do not actually build them by themselves.

Two companies you have probably never heard of are the secret winners. Broadcom and Marvell together control an estimated 95% of the custom AI chip co-design market, serving as the engineering partners that translate the hyperscalers' chip ideas into manufacturable silicon. These two firms do the deep engineering work that turns Google's or Amazon's chip concept into something a factory can actually produce. As the custom chip boom grows, they are cashing in. Broadcom alone carries a $73 billion AI backlog and is targeting $100 billion in annual AI chip revenue by 2027. Den of GeekFremontleaf

And there is one company that every single player depends on, including Nvidia. All of this is enabled almost entirely by TSMC, which fabricates the chips for all five hyperscalers and for Broadcom. The Taiwanese manufacturer TSMC physically makes nearly all the advanced chips in the world, whether they are designed by Nvidia, Google, or Amazon. TSMC produces approximately 92% of advanced AI chips. No matter who wins the design battle, TSMC builds the actual hardware, which is part of why the tiny island of Taiwan sits at the center of the entire global technology economy. FremontleafChevrolet Philippines

Tom's Hardware on the custom chip state of play: https://www.tomshardware.com/tech-industry/semiconductors/custom-ai-asics-examined-from-broadcom-to-mtia

That investor was wrong. Gamma is now worth $2B, with 50M users and more than half their growth driven by word of mouth.

They're one of 6 AI-native startups in HubSpot for Startups' free Bold Bets Playbook. Replit grew revenue 50x after half the team pushed back on the strategy. Ramp generated 100M+ views from a single stunt. Clay's co-founder wouldn't hang up a sales call until the prospect DMed him in Slack.

Each one took a GTM risk most founders would never greenlight. Each one paid off.

Get the free Bold Bets playbook

You do not buy AI chips, so why should you care? A few reasons.

First, it could make AI cheaper for everyone. The reason these companies are building their own chips is to slash the cost of running AI. As we covered recently, the era of cheap, subsidized AI is ending and companies are starting to charge for what you use. Custom chips are the other side of that story: if Google, Amazon, and Microsoft can run AI far more cheaply on their own hardware, some of those savings could eventually reach you through lower prices on the AI tools and cloud services you use.

Second, it explains the strange intensity of the chip world right now. The hundreds of billions being spent, the obsession with Taiwan and TSMC, the rise of companies like Broadcom, all of it traces back to this single dynamic: everyone is racing to control the chips that power AI, because whoever controls the chips controls the cost and the future of the entire industry.

Third, it is a reminder that even the most dominant companies are never as safe as they look. Nvidia looks unstoppable today, with sales off the charts and a place at the center of the AI boom. But its own best customers are spending billions to need it less. That is how technology works. Today's indispensable giant is tomorrow's negotiating partner, and the quiet moves being made right now in chip design will shape which companies dominate the next decade. The AI you use every day runs on a battle most people never see, and that battle is heating up fast.

We will keep tracking the chip wars and bring you the next chapter as it lands. Stay sharp out there.