Nvidia leads $1 trillion sales race as AI chip rivals close in

Three companies are racing toward $1 trillion in annual sales, a milestone no corporation has ever reached.

Nvidia, Amazon, and SpaceX are racing toward $1 trillion in annual sales, a milestone no company has reached. Nvidia leads with $92 billion in quarterly revenue, nearly double a year earlier.

"We're quite excited about what's happening in our chips business," Amazon CEO Andy Jassy told analysts in July, noting the company's semiconductor operation would generate more than $25 billion in annual revenue as a standalone business, with sales growing at triple-digit rates.

Nvidia's earnings report on Wednesday showed revenue nearly doubled to $92 billion, with management guiding to 70 percent growth next year. The company retains roughly 90 percent of the AI accelerator market, charging about $30,000 per chip with profit margins near 75 percent. But the field is expanding: AMD's data center revenue more than doubled to $6.7 billion last quarter, Broadcom expects to sell $56 billion of AI products this year, and 150 companies are developing more than 200 AI chip designs, according to Jon Peddie Research.

The AI chip market is expected to cross $1 trillion in revenue, and the economics are shifting as data centers move from training models to inference. Nvidia's $5.08 trillion market cap and forward P/E near 16x suggest the market still prices in dominance, but Amazon's in-house chip business growing at triple-digit rates and OpenAI's custom silicon deployment later this year could reshape the economics.

The Challengers Are Multiplying

Nvidia's dominance in AI accelerators has made it the world's most valuable company, but its biggest customers are becoming its fiercest competitors. Amazon, Google, and Meta are all designing custom chips, often with help from Broadcom and Marvell. Broadcom expects to sell $56 billion of AI products this year, with AI-related revenue at its design division more than doubling, according to Bloomberg Intelligence.

Google is working with Marvell on its tensor processing units and will release two versions simultaneously this year for the first time. OpenAI's in-house chip, called Jalapeno, developed with Broadcom, will be deployed to data centers later this year, initially focused on inference. Anthropic has lined up tens of billions of dollars in contracts to use AI accelerators from Google, Amazon, and AMD.

Startups are also entering the fray. Cerebras held the semiconductor industry's largest-ever initial public offering in May, and its CEO Andrew Feldman pitches chips that are much faster than Nvidia's at responding to AI prompts. "This is a market that cares desperately about speed," he said. Other startups like Positron, Fractile, and Etched are attracting rapidly growing valuations.

Inference Opens the Door

The shift from training AI models to inference — the stage when models respond to real-world inputs — is creating opportunities for a wider range of processors. Nvidia's GPUs are powerful for training, but inference workloads can be handled more efficiently by specialized chips. "Inference is not a one-size-fits-all, so brute-forcing inference with a single chip is not going to work," said Sid Sheth, CEO of d-Matrix, a startup focused on the space.

Nvidia is not standing still. The company agreed to pay a reported $20 billion for technology and personnel from Groq, a startup focused on inference, and has agreed to buy Hugging Face for $12.9 billion. It is also overhauling its chip designs every year and expanding into data center cooling, networking, and storage.

The supply chain remains a critical constraint. Taiwan Semiconductor Manufacturing Co. produces the most advanced chips, and Nvidia has locked in supply commitments above $100 billion, according to Bloomberg Intelligence analyst Kunjan Sobhani. "As AI-infrastructure spending accelerates, Nvidia's size and purchasing power can become competitive advantages themselves," he said.

For investors, the race to $1 trillion in sales is also a race for market share in the AI economy. Nvidia shares trade at roughly 16x next year's consensus earnings, a discount that some analysts argue is unjustified given 70 percent growth guidance. But the competitive pressure is real: Amazon still sends about a quarter of its capital expenditures to Nvidia, yet its in-house chip business is growing at triple-digit rates. The question is how long Nvidia can maintain its 90 percent share of the accelerator market as hyperscalers, AI labs, and startups all target the same prize.

This article is for informational purposes only and does not constitute investment advice.