In short
Nvidia is still dominant in AI accelerators, but WIRED says its lead is under pressure from rivals, custom chips, AI-driven portability and China restrictions. The story also highlights a leadership shift as Jensen Huang steps back and Colette Kress takes over day-to-day control.
- Nvidia still holds about 80% of AI accelerator sales, but its lead is facing broader pressure.
- AMD’s AI accelerator share rose from 2% in 2022 to 7%-10% in 2026, according to WIRED.
- China’s share of Nvidia revenue has fallen to 8% from 20% in 2024 because of U.S. export controls.
- Google, Amazon and Microsoft are building custom chips to cut costs and dependence on Nvidia.
- AI tools are making it easier to port models away from Nvidia’s CUDA ecosystem.
Nvidia is entering a more complicated phase just as its chief executive, Jensen Huang, is stepping out of day-to-day control. WIRED says the company still commands roughly 80% of AI accelerator sales, but that lead is being squeezed by rivals, custom chips from cloud giants, and U.S.-China tensions that have already cut China’s share of Nvidia revenue to 8% from 20% in 2024.
The leadership shift matters because it lands at the same moment Nvidia’s market is broadening beyond training chips into inference, where AI agents, software portability and general-purpose processors could weaken the company’s long-standing advantage.
What changed at Nvidia?
According to WIRED, Huang is exiting the company’s day-to-day operational control, with Colette Kress taking over that role. The story does not suggest Huang is disappearing from the company entirely, but it does frame the transition as a major inflection point for a business that has been shaped by Huang’s long tenure and his willingness to place early bets on AI.
Huang founded Nvidia in 1993, when the company was focused on graphics chips for PCs. Over time, he pushed it toward a far broader AI role, backing research funding, software tooling and hardware for labs and startups. That early positioning helped Nvidia become the dominant supplier for AI training hardware through its GPUs and CUDA software stack.
Now, the question is whether the same playbook can hold in a market where customers have more alternatives and the next wave of AI demand looks different from the last one.
Why it matters
This is not just a boardroom transition. Nvidia’s moat is being tested from several directions at once. WIRED reports that competitors are taking share, cloud companies are building their own silicon, and AI tools are making it easier to move models off Nvidia’s stack and onto rival hardware. At the same time, export restrictions are limiting Nvidia’s access to China, while that market remains important enough to affect the company’s overall growth trajectory.
For investors, the issue is less whether Nvidia stays large than whether it can keep growing at the pace implied by its recent revenue surge. WIRED says revenue climbed from $26.91 billion in January 2022 to $215.94 billion in January 2026. That kind of rise is hard to repeat if the market shifts from a single dominant supplier to a more fragmented mix of hyperscaler chips, startup hardware and alternative architectures.
For customers, the stakes are cost and flexibility. As Google, Amazon and Microsoft build custom chips to reduce dependence on Nvidia, they are also signaling that scale buyers want more control over pricing and supply. If that trend continues, Nvidia may have to defend its margins more aggressively than before.
How the AI shift changes the chip market
AI training and AI inference are not the same business. Training requires enormous compute to build models. Inference is what happens after a model is trained: it answers queries, runs agents and powers deployed applications. WIRED argues that the rise of AI agents shifts value toward inference, a segment where hyperscalers and startups may have structural advantages.
That matters because Nvidia’s dominance was built first on training. The company is still strong there, but inference opens the door to different kinds of hardware and software tradeoffs. If the fastest-growing workloads are no longer the ones Nvidia best controls, its strategic edge becomes harder to protect.
How rivals are closing in
AMD is the clearest challenger in the material provided. WIRED says AMD’s share of the AI accelerator market rose from 2% in 2022 to a range of 7% to 10% in 2026. That is still far behind Nvidia, but it signals that the market is no longer a one-company story.
Big cloud providers are also part of the competitive pressure. Google, Amazon and Microsoft have all been building custom chips to cut costs and reduce dependence on Nvidia, according to WIRED. Their incentive is straightforward: if they can design enough of their own infrastructure, they can lower spending and reduce supply risk.
Then there are the smaller specialists. WIRED says specialized startups are targeting niches Nvidia covers less well, especially inference. These companies are not trying to dethrone Nvidia all at once. Instead, they are carving out the parts of the market where Nvidia’s broad platform may be less tailored.
Can code portability weaken CUDA?
Yes, and that is one of the more important shifts in the story. WIRED says AI tools are making it faster to port models from Nvidia’s CUDA stack to rival hardware. In other words, the software lock-in that helped Nvidia for years may be loosening.
That does not mean migration is effortless. Nvidia still benefits from installed base, developer familiarity and the sheer scale of its ecosystem. But if moving workloads becomes cheaper and faster, competitors do not need to match Nvidia feature-for-feature to win business. They only need to make switching more attractive.
“They have access to AI too,”
Patrick Moorhead, chief analyst at MoorInStrat, told WIRED in a comment that captures the broad trend: rivals and major customers are no longer locked out of AI capability. They can build, buy or adapt their own paths forward.
China is the other pressure point
China remains central to Nvidia’s story, even as it becomes less lucrative. WIRED says China now accounts for 8% of Nvidia revenue, down from 20% in 2024. U.S. export controls are blocking Nvidia’s most advanced chips from the Chinese market, though older chips can still be sold there.
That change hurts on two levels. First, it removes a direct source of premium sales. Second, it may push Chinese companies to accelerate domestic alternatives. WIRED notes that Huawei is developing large-chip-count systems for AI training and inference, and that Z.ai says it trained models entirely on Huawei chips.
Those developments matter because they suggest China is not merely waiting out export controls. It is building a parallel supply chain, one that could mature into a more durable competitive threat.
What do older chips still mean for Nvidia?
Older chips still matter because they keep some revenue flowing into China, but they do not fully solve the strategic problem. WIRED says sales of older China-legal chips and non-AI chips doubled over the past year, which suggests Nvidia is finding ways to sell around restrictions. Yet those are not the same as the company’s most advanced products.
That split leaves Nvidia in a difficult position: it can preserve some commercial access, but not necessarily the most profitable or strategically important kind.
| Metric | Earlier figure | Latest figure | What it suggests |
|---|---|---|---|
| AMD share of AI accelerators | 2% in 2022 | 7%-10% in 2026 | Competitor momentum is building |
| Nvidia revenue | $26.91 billion in January 2022 | $215.94 billion in January 2026 | The company grew at extraordinary speed |
| China’s share of Nvidia revenue | 20% in 2024 | 8% now | Export controls are reshaping the business |
| Megaspeed chip imports | $4.6 billion-$7.5 billion reportedly imported into southeast Asia | Indirect chip flow may be under scrutiny | |
| Internal employee comments | 1,600+ comments on a suggestion-box post | Leadership uncertainty triggered internal reaction | |
Background: how Huang built the company
Huang’s significance to Nvidia is hard to overstate. He co-founded the company in 1993 and spent years steering it through a period when it was primarily known for graphics chips for PCs. The decisive pivot came in the mid-2010s, when he leaned into AI chips and the software ecosystem needed to support them.
That bet paid off spectacularly after ChatGPT mainstreamed AI in 2022. Demand for Nvidia hardware surged as model builders raced to secure training capacity. In effect, Nvidia became the default supplier for the boom.
WIRED says Huang backed AI early not only with chips but with research funding, software tooling and hardware for labs and startups. That combination helped make Nvidia more than a component vendor. It became the platform on which much of the AI industry was built.
But platforms eventually face substitutes. Once the market is mature enough, buyers begin asking whether one vendor should control so much of the stack.
Where the competition is coming from now
WIRED’s broader argument is that Nvidia’s challenge is no longer only about a single rival. It is about an ecosystem moving away from dependence on one company.
1. Hyperscalers are making their own silicon
Google, Amazon and Microsoft have the scale and balance sheets to support custom chip programs. Their goal is to lower costs and reduce Nvidia dependence. Even partial success could take a meaningful slice out of Nvidia’s addressable market.
2. Startups are targeting specific workloads
Specialized startups are focusing on the parts of the stack Nvidia covers less well, especially inference. That is where speed, power efficiency and software flexibility can matter more than raw training muscle.
3. AMD is gaining ground
AMD’s rise from 2% to 7%-10% does not make it the market leader, but it does show that customers are experimenting. Once large buyers standardize on more than one supplier, the market becomes more competitive, and pricing power can slip.
4. AI is making hardware switching easier
Porting has historically been one of Nvidia’s biggest advantages. If AI tools can translate workloads faster, the moat narrows. WIRED says Wafer can get new open-weight models running on AMD hardware at least 80% as efficiently within hours, underscoring how quickly interoperability can improve.
Leadership risk and the Intelligencer angle
WIRED focuses on market and strategic pressures, but Intelligencer framed Huang’s departure in more dramatic terms, calling it part of a broader scandal involving sex, espionage and investor panic. That outlet also said internal employee comments pushed for Huang’s return after his departure.
One anonymous commenter on the internal suggestion box wrote, “If you believe in the mission, bring back THE boss”. Intelligencer said the post drew more than 1,600 comments, a sign that the leadership question was not just external theater. It was also an internal morale issue.
The two outlets are looking at different parts of the same story. WIRED emphasizes business pressure and changing market structure. Intelligencer emphasizes governance, personal controversy and the emotional reaction inside the company. Taken together, they suggest Nvidia’s leadership change is landing at an unusually sensitive moment.
What the numbers say about Nvidia’s position
Nvidia is still enormous, but the numerical picture is more mixed than the company’s headline growth once suggested.
- 80% of AI accelerator sales still flow through Nvidia, according to WIRED.
- $215.94 billion is the revenue figure WIRED cites for January 2026, up from $26.91 billion in January 2022.
- 8% of Nvidia revenue now comes from China, down from 20% in 2024.
- 2% was AMD’s AI accelerator share in 2022; by 2026, it had climbed to 7%-10%.
- $4.6 billion to $7.5 billion is the amount Megaspeed and subsidiaries reportedly imported into southeast Asia in Nvidia chips, according to WIRED.
Those figures tell a story of extraordinary scale, but also of a market that is becoming harder to control from the center.
What happens next?
The most immediate question is how much authority Colette Kress will actually have and whether Huang’s withdrawal from day-to-day control marks a temporary adjustment or a lasting handoff. WIRED does not answer that definitively, and neither does the broader fact sheet.
Other questions are more commercial than personal. How much of Nvidia’s China-related chip flow reaches Chinese buyers indirectly? Can rival chips and AI-based code translation really erode CUDA’s lock-in? And how fast can hyperscalers and startups scale custom silicon beyond niche deployments?
There is also the governance question. Intelligencer’s reporting suggests that employee sentiment and controversy around Huang may continue to shape the narrative. If additional allegations or disclosures emerge, the leadership story could become even more consequential for investors and customers.
What to watch in the months ahead
- Whether Kress’s control is operational only, or a sign of a broader succession plan.
- Whether AMD’s share keeps rising or stalls in the high single digits.
- Whether cloud providers push custom chips deeper into mainstream use.
- Whether Huawei and other Chinese efforts keep improving enough to reduce dependence on Nvidia.
- Whether direct or indirect China sales continue to offset export restrictions.
For now, Nvidia remains the defining company of the AI hardware boom. But the next phase looks less like a one-company race and more like a contest over who owns training, inference, software compatibility and supply chains. That is a much harder environment for any incumbent, even one as dominant as Nvidia.
Timeline
| Year | Event |
|---|---|
| 1993 | Jensen Huang co-founds Nvidia |
| Mid-2010s | Huang pivots the company toward AI chips and supporting software |
| 2022 | ChatGPT mainstreams AI and Nvidia chip sales surge |
| 2024 | China accounts for 20% of Nvidia revenue, per WIRED |
| Dec. 2025 | Nvidia acquires specialized-chip startup Grok |
| Jan. 2026 | Nvidia revenue reaches $215.94 billion, according to WIRED |
| 2026 | AMD share of the AI accelerator market rises to 7%-10% |
| Nov. 13, 2026 | Intelligencer publishes its roundup on the Huang scandal coverage |
| Nov. 16, 2026 | WIRED publishes its analysis of Nvidia’s competitive and China pressures |
Frequently asked questions
Why is Nvidia under pressure now?
Nvidia is under pressure because multiple forces are converging at once: competitors are gaining share, cloud giants are building their own chips, AI tools are making model portability easier, and U.S. export controls have reduced access to China. WIRED argues those pressures could slow future growth and weaken margins.
How much of the AI chip market does Nvidia still control?
Nvidia still controls about 80% of AI accelerator sales, according to WIRED. That remains a dominant position, but it is no longer as insulated as it once was because rivals, hyperscalers and specialized startups are all pushing into parts of the market.
What does Colette Kress taking over day-to-day control mean?
It means Jensen Huang is stepping back from daily operations, with Colette Kress handling the company’s routine leadership. WIRED does not say whether this is temporary or permanent, so the key question is how much authority she will ultimately have.
Is China still important to Nvidia?
Yes, but less than before. WIRED says China now accounts for 8% of Nvidia revenue, down from 20% in 2024. Export controls limit sales of Nvidia’s most advanced chips there, though older chips still appear to be sold into the market.
Are Nvidia rivals really catching up?
They are catching up in parts of the market, not overall dominance. WIRED says AMD’s share rose from 2% in 2022 to 7%-10% in 2026, while Google, Amazon and Microsoft are building custom chips and startups are targeting inference. That does not dethrone Nvidia, but it does narrow its moat.









