A rumor that Nvidia (NVDA) was in talks to acquire a PC manufacturer stirred up capital market waves a week ago. On April 13, tech news website SemiAccurate claimed that Nvidia had been negotiating to acquire a large PC company for more than a year, a deal that would “reshape the personal computer and server landscape.”
Although it did not mention names, the market quickly fixed its gaze on Dell and HP, the world’s second and third largest PC shippers. Dell’s stock surged 6.74% to close at $189.79, a record high on the day; HP also rose 5.3% to $19.23. Dell’s market capitalization briefly touched $122.6 billion during trading.
However, an Nvidia spokesperson quickly denied the report, stating it was “untrue” and that the company was not engaged in any discussions to acquire a PC maker. In after-hours trading, Dell and HP shares fell more than 2% in response, bringing the brief frenzy sparked by the rumor to an end.
An illogical move – why did investors still “buy it”?
From a business logic perspective, such a deal is almost impossible. Nvidia’s gross margin has long remained in the 70%–75% range, while Dell’s is only around 22%, and HP’s is less than 20%. Gene Munster, managing partner of Deepwater Asset Management, said bluntly: “Nvidia is still navigating its gross margin opportunity. Acquiring a PC maker would be a headwind to navigate.” He rated the probability of such a deal as “less than 50%.”
So why did the market still believe this seemingly absurd story?
In fact, the sharp volatility triggered by this rumor – “Dell and HP shares first rose, then fell” – precisely illustrates that the logic behind it has strong “narrative plausibility” in the minds of investors. This plausibility is not based on simple M&A imagination, but on the deep-seated logic of the AI industry’s transition from a technological explosion phase to an ecosystem harvesting phase. The market is betting on an ultimate script: Nvidia wants to evolve from a “pick-and-shovel seller” into a “rule definer,” replicating the Apple model through vertical integration, completely shaking off its dependence on downstream OEM manufacturers, and directly defining the standards for the next generation of AI personal computing devices.
The truly dangerous signal: the inference layer is undergoing “de-Nvidia-fication”
Putting aside the truth of the acquisition itself, the rumor reveals an accelerating industry trend – Nvidia’s moat in the cloud is not unassailable. Take inference, the fastest-growing segment of AI computing. After a model is trained, every user query and every AI response is inference. And it is precisely in this area that hyperscale cloud providers are systematically reducing their reliance on Nvidia.
Google’s self-developed TPU has been iterated to the seventh generation, Ironwood, designed specifically for inference, with ten times the performance of its predecessor, and is now being leased at scale to customers. Amazon’s Trainium 2 chip is being used by Anthropic to train large models, with a scale of 500,000 chips. Microsoft’s Maia accelerator and Meta’s MTIA chip are also being continuously iterated. These tech giants were strangled by Intel a decade ago, and they don’t want Nvidia to repeat that performance.
And the technology backend supporting all this is precisely Broadcom (AVGO). Broadcom dominates the backend design services for custom AI accelerator chips (ASICs). Its AI revenue last quarter reached $8.4 billion, a year-over-year surge of 106%. According to Counterpoint Research, Broadcom will control 60% of the custom AI chip market next year. As Nvidia’s major customers turn to custom chips, Broadcom becomes the most critical “arms dealer” in this “hunt.”
Meanwhile, another American company, Marvell (MRVL), is also trying to get a piece of this pie. On April 20, The Information reported that Google is in talks with Marvell to develop two new chips – a memory processing unit to supplement TPUs, and a new TPU specifically designed for inference. After the news broke, the company’s shares jumped 7% in pre-market trading.
Nvidia’s real counterattack: edge AI, not acquiring a PC maker
Nvidia is clearly aware of its predicament. The logic behind the rumor seems plausible precisely because Nvidia does need to find its next battlefield – and that battlefield is most likely edge AI computing.
Nvidia has confirmed that it will launch dedicated AI PC chips, the N1 and N1X, in the second half of 2026. These chips will use the ARM architecture, TSMC’s 3nm process, and natively support the Windows on ARM operating system. The chip is being co-developed by Nvidia and MediaTek, optimized for high-efficiency edge AI tasks. This move aims to build a unified AI computing ecosystem covering both enterprise deployments and personal computing scenarios, directly responding to the strategic goal implied by the “PC maker acquisition” rumor – taking control of the AI computing entry point on end devices.
If successful, Nvidia will not need to acquire Dell or HP to continue occupying a core position in the edge computing era. In short, the reason a rumor denied by the company could trigger billions of dollars in market cap volatility is not because the message itself was so reliable, but because it touched the market’s deepest cognitive anxiety – Nvidia’s moat at the data center inference layer is being eroded by custom chips. Regardless of whether any acquisition happens, one basic fact has already been established: Nvidia’s dominant position in the cloud is not unchallengeable, and the battle for edge AI computing has already quietly begun.
