AMD is Nvidia’s biggest direct competitor in AI chips. It sells rival data center GPUs that are now deployed by Anthropic, Meta, Microsoft, OpenAI, and Oracle. In the long run, Nvidia’s biggest threat may be its own customers: Google, Amazon, and Microsoft now design their own AI chips, with help from Broadcom and Marvell.
Nvidia is still far ahead. It earned $96.2 billion in its latest quarter, and $89 billion of that came from data centers. Below, you will find Nvidia’s main competitors in AI chips, custom silicon, start-ups, and China, plus why Nvidia’s CUDA software keeps it ahead.
Key Takeaways
- AMD is Nvidia’s closest direct rival, with data center sales of $6.7 billion in its latest quarter, up 107%.
- Nvidia’s data center revenue was $89 billion in the same period, more than 13 times AMD’s.
- Google, Amazon, and Microsoft build their own AI chips, turning major Nvidia customers into competitors.
- Broadcom’s AI chip revenue jumped 221% to $16.7 billion in one quarter, mostly from custom chips for big tech.
- In China, Huawei has cut Nvidia’s share of data center AI chips to 34%, according to IDC.
Who Are Nvidia’s Main Competitors?
Nvidia competes with four groups of companies. Here is how they compare:
| Competitor | Type | Main AI Chip | Latest Data Point |
|---|---|---|---|
| AMD | GPU maker | Instinct MI400 series | Data center sales $6.7B, up 107% |
| Intel | Chipmaker | Gaudi 3, Crescent Island | New inference GPU sampling in late 2026 |
| Google (Alphabet) | Big tech, own chips | TPU (Ironwood) | Selling TPUs to outside customers in 2026 |
| Amazon (AWS) | Big tech, own chips | Trainium | Custom chip business at a $20B+ yearly pace |
| Microsoft | Big tech, own chips | Maia 300 | Announced October 2026 |
| Broadcom | Custom chip designer | Custom AI accelerators | AI revenue $16.7B, up 221% |
| Marvell | Custom chip designer | Custom silicon, networking | Data center revenue up 46% |
| Cerebras | Start-up | Wafer-scale chip | IPO raised $5.55B in May 2026 |
| Huawei | Chinese chipmaker | Ascend | 28% of China’s data center AI chips |
Sources: latest quarterly reports and company announcements, 2026.
Nvidia’s own annual report names many of these companies as competitors, including AMD, Huawei, Intel, Alphabet, Amazon, Microsoft, Broadcom, and Marvell.
Who Is Nvidia’s Biggest Competitor?
AMD is Nvidia’s biggest competitor today. It sells data center GPUs and full rack systems that compete head to head with Nvidia’s, and it is the first rival Nvidia names in its annual report.
AMD’s revenue grew 50% to $11.5 billion in the second quarter of 2026, and data center sales made up 58% of the total, according to its quarterly report. Its new Helios rack systems are being deployed by Anthropic, Meta, Microsoft, OpenAI, and Oracle. Anthropic alone agreed to deploy up to 2 gigawatts of AMD’s MI450 chips.
Still, the gap is large. Nvidia’s data center revenue of $89 billion in its latest quarter was more than 13 times AMD’s, according to Nvidia’s second-quarter results.
Who is Nvidia’s newest rival? The newest serious challengers are Nvidia’s own customers. In October 2026, Axios described the trend as “the AI chip war is turning Nvidia’s customers into rivals,” after Microsoft revealed its Maia 300 chip.
Which GPU Makers Compete With Nvidia?
1. AMD
AMD, led by CEO Lisa Su, launched its Instinct MI400 series in 2026, including the MI455X and MI430X. It sells complete rack systems, called Helios, to compete with Nvidia’s full data center systems. AMD expects its data center sales to grow faster in the second half of 2026.
2. Intel
Intel sells Gaudi 3 AI accelerators and is preparing Crescent Island, a new data center GPU built for AI inference with 160GB of memory. Customer samples are planned for the second half of 2026. Intel is also a partner: Nvidia agreed in September 2025 to invest $5 billion in Intel for about a 4% stake, and the two are building chips together. Read more in our guide to Intel’s competitors.
How Are Big Tech Companies Competing With Nvidia?
Big tech companies buy huge numbers of Nvidia chips, but they now design their own chips too. Google Cloud and Microsoft Azure are among the partners running Nvidia’s newest Vera Rubin systems, even as both build in-house alternatives.
3. Google
Google designs Tensor Processing Units (TPUs). Its newest, Ironwood, launched in November 2025, and customers can link up to 9,216 Ironwood chips in one pod. In April 2026, Google said it would sell TPUs to a select group of outside customers for their own data centers. Anthropic signed a deal for 3.5 gigawatts of TPUs, and Meta has reportedly agreed to rent TPUs through Google Cloud. See how Google competes more widely in our list of Google’s competitors.
4. Amazon
Amazon Web Services makes Trainium chips for AI. Amazon’s custom chip business passed a $20 billion annual revenue pace in 2026 and is growing more than 100% a year, according to The Register. Anthropic committed to up to 5 gigawatts of Trainium chips, and OpenAI committed to about 2 gigawatts starting in 2027. Our guide to Amazon’s competitors covers its other rivals.
5. Microsoft
On October 12, 2026, Microsoft announced Maia 300, its newest in-house AI chip for inference, the work of running AI models for users. Microsoft claims it produces 40% more output per megawatt than leading GPU systems, a figure that has not been independently verified. Microsoft still uses Nvidia chips for training and plans to deploy Nvidia’s Vera Rubin systems. Learn more about Microsoft’s competitors.
6. Meta
Meta buys chips from several suppliers at once. It is deploying AMD’s Helios racks and reportedly renting Google’s TPUs. Every chip Meta buys from someone else is a sale Nvidia does not make. For Meta’s own business rivals, see our list of Facebook competitors.
Who Helps Big Tech Build Custom Chips?
7. Broadcom
Broadcom designs custom AI accelerators for big tech companies. Its AI chip revenue rose 221% to $16.7 billion in the quarter ended August 2, 2026, and it expects $21.7 billion in the next quarter. CEO Hock Tan said demand for its “custom AI accelerators and networking continues to be very strong.” Broadcom will also help develop Google’s TPUs through 2031.
8. Marvell
Marvell makes custom chips and networking parts for data centers. Its data center revenue grew 46% in its latest quarter, and it expects its custom chip business to speed up in the second half of its fiscal year.
Which AI Chip Start-Ups Compete With Nvidia?
9. Cerebras
Cerebras builds very large “wafer-scale” chips for AI. It went public in May 2026, raising $5.55 billion at $185 a share, and its stock closed its first day at $311, giving it a value of about $95 billion. OpenAI signed a $10 billion deal to use about 750 megawatts of Cerebras computing power through 2028.
10. Groq
Groq makes chips designed for fast AI inference. In December 2025, Nvidia agreed to pay $20 billion to license Groq’s technology and hire its founder, Jonathan Ross, and other leaders. Groq remains an independent company and still runs its GroqCloud service.
Why would Nvidia pay $20 billion for a rival’s technology? Inference is a growing battleground: Microsoft’s Maia 300 and Intel’s Crescent Island are both built for it. Nvidia’s CEO Jensen Huang said the company plans to add Groq’s low-latency processors to its AI factory systems, without buying Groq as a company.
Who Competes With Nvidia in China?
11. Huawei
Huawei is Nvidia’s biggest rival in China. In the first half of 2026, Nvidia held 34% of China’s data center AI chip market and Huawei held 28%, according to IDC data reviewed by Reuters. Nvidia’s shipments there fell 44%, partly because of US export limits and China’s push for local chips. Huawei plans to release its Ascend 960DT chip in early 2027.
Nvidia’s own forecast for its current quarter assumes no data center computing revenue from China.
Why Is Nvidia Still Ahead?
Nvidia’s biggest advantage is not just its chips. It is CUDA, the software platform developers use to program Nvidia chips. More than 7.5 million developers use CUDA and Nvidia’s other software tools, according to Nvidia’s annual report.
This is a classic example of an ecosystem moat. Once companies build their AI software on CUDA, switching to another chip means rewriting and retesting code. That makes customers slow to leave, even when a rival chip is cheaper. It is the same reason a strong market position is hard to attack: the leader’s advantage grows with every customer.
Can Anyone Overtake Nvidia?
Not soon. Nvidia’s revenue more than doubled to $96.2 billion in its latest quarter, and it expects about $108 billion in the next one. Its gross margin of 75% also shows customers still pay a premium for its chips.
The real risk is slower and quieter. As Google, Amazon, and Microsoft move more of their own work to in-house chips, Nvidia’s biggest customers will buy a smaller share of their chips from Nvidia. Microsoft plans multi-gigawatt Maia deployments starting in 2027, and Amazon says it has more than $225 billion in Trainium revenue commitments.
Nvidia Competitor Stocks
For readers following the stock market, here are the publicly traded rivals mentioned above:
| Company | Ticker | Exchange |
|---|---|---|
| Nvidia | NVDA | Nasdaq |
| AMD | AMD | Nasdaq |
| Intel | INTC | Nasdaq |
| Broadcom | AVGO | Nasdaq |
| Marvell | MRVL | Nasdaq |
| Microsoft | MSFT | Nasdaq |
| Cerebras | CBRS | Nasdaq |
This table is for information only and is not investment advice.
Bottom Line
AMD is Nvidia’s biggest competitor today, but the bigger long-term threat comes from Nvidia’s own customers. For me, this is one of the most interesting rivalries in business: the companies spending the most on Nvidia’s chips are the same ones building their way out. Nvidia’s answer is CUDA, new platforms like Vera Rubin, and deals like Groq. For now, the numbers still favor Nvidia by a wide margin.
Frequently Asked Questions
Nvidia is far larger in AI chips today, with $89 billion in quarterly data center sales versus AMD’s $6.7 billion. AMD is growing fast and wins major customers like Anthropic, Meta, and OpenAI, often by offering an alternative to Nvidia’s systems.
Yes. Google designs its own AI chips, called TPUs, and began selling them to select outside customers in 2026. Anthropic signed a deal for 3.5 gigawatts of TPUs, and Meta has reportedly agreed to rent them through Google Cloud.
Yes, but it is also a partner. Intel sells Gaudi AI chips and is preparing a new inference GPU called Crescent Island. At the same time, Nvidia agreed to invest $5 billion in Intel in 2025, and the two companies are building chips together.
Companies most similar to Nvidia are other AI chipmakers: AMD, Intel, Broadcom, Marvell, and newer firms like Cerebras. Big tech companies such as Google, Amazon, and Microsoft also design their own AI chips and offer them through their cloud services.
Only in limited amounts. US export rules and Chinese government policy have cut Nvidia’s sales there. Its share of China’s data center AI chips fell to 34% in early 2026, and its current forecast assumes no data center computing revenue from China.
