Nvidia’s Record Quarter Meets a Changing AI Landscape: Rubin CPX Cancelled, Vera CPU Rises, and Rotation Weighs on the Stock
27.05.2026 - 22:21:02 | boerse-global.de
Nvidia just delivered the strongest quarter in its history, yet the stock cannot catch a bid. The shares slipped to €181.32 on Tuesday, down 1.7%, extending a weekly slide of 5.63%. The relative strength index sits at 39.5, flirting with oversold territory. That is an unusual place for a company that grew revenue 85% to a record $81.62 billion.
Two forces are pulling the stock in opposite directions. One is a rotation within the semiconductor sector itself: investors are rewarding memory and networking names at the expense of GPU pure plays. The other is a quiet overhaul of Nvidia’s product roadmap — a cancellation, a licensing deal, and an in-house CPU that together signal the company is trying to break free of its dependence on a single chip architecture.
The roadmap gets a rethink
The clearest sign of change is the fate of the Rubin CPX, an inference chip that had been slated to pack 128 GB of GDDR7 memory and arrive in the second half of this year. Supply-chain signals have dried up: orders for the required memory and specialised substrates are absent. At the GTC 2026 conference, the chip did not appear on Nvidia’s roadmap at all. In industry circles, the project is now considered effectively cancelled, even if the company is not calling it a retreat.
Instead, Nvidia has moved to license technology from Groq 3 LPX in a deal worth $20 billion. The emphasis is shifting from raw GPU performance to lower latency, real-time inference, and specialised AI infrastructure. The shift mirrors a broader change in how Nvidia wants to position itself: less a component vendor, more an integrated platform company.
Should investors sell immediately? Or is it worth buying Nvidia?
Agentic AI becomes an industrial lever
That platform ambition is visible in a new push around “agentic AI”. At a Rescale event, Nvidia unveiled PhysicsNeMo as the core of a platform for agent-based digital development. The idea is that AI agents take over parts of complex simulations, accelerating engineering workflows. Early data from McLaren Automotive suggests the payoff could be substantial: a threefold productivity gain, a 30-fold improvement in cost efficiency, and a 60% reduction in development cycles.
For industrial customers, the appeal is concrete. The same time window can now cover four times as many design candidates. This is not another AI buzzword; it is a measurable return on investment that Nvidia is banking on to open new revenue streams beyond the datacentre.
Vera CPU benchmarks turn heads
On the hardware side, Nvidia is not abandoning silicon. Its first internally developed CPU, codenamed Vera, has posted strong Linux benchmark results. The processor uses 88 Olympus ARM cores and delivers 1.2 TB/s of memory bandwidth. Against Intel’s Xeon 6980P, it is 55% faster; against AMD’s EPYC 9575F, the edge is 10%.
The architecture targets significantly better energy efficiency than traditional x86 designs. Analysts put the addressable market for the Vera platform at $200 billion, and internal projections see CPU-related revenue reaching $20 billion by the end of fiscal 2027. If that plays out, Nvidia will control not just the GPU that runs AI workloads but the central processor that orchestrates the server.
Record numbers lose their sting
All this strategic manoeuvring is being financed from a position of extraordinary strength. In the first quarter of fiscal 2027, the datacentre business alone generated $75.25 billion, up 92% year on year. Adjusted earnings per share climbed from $0.76 to $2.39, and the gross margin hit 74.9%. Management has guided for around $91 billion in second-quarter revenue, though that figure excludes any contribution from China.
Despite these numbers, the market is preoccupied with where the next wave of AI spending will land. The Philadelphia Semiconductor Index reached an all-time high, but Nvidia was left behind. Investors are rewarding makers of high-bandwidth memory and networking gear — the components that support GPUs rather than the GPUs themselves.
Nvidia at a turning point? This analysis reveals what investors need to know now.
Buyback and dividend fail to lift the mood
Nvidia has tried to soothe shareholders with capital returns. The board authorised an $80 billion share buyback and increased the quarterly dividend from $0.01 to $0.25 per share. The signal was clear, but the market ignored it. The stock now sits roughly 9% below its 52-week high of €201, still up 53% over twelve months and 13% year to date.
The question hanging over Nvidia is not whether AI demand exists — it clearly does, with outstanding purchase commitments of $119 billion binding the supply chain. The question is how that demand will be distributed across the infrastructure stack. Nvidia’s partnership with Marvell Technology, involving a $2 billion investment in custom XPUs and networking for NVLink Fusion, suggests it wants to keep controlling the architecture end to end.
The cancellation of the Rubin CPX and the rise of Vera, along with the Groq licensing deal and the agentic AI push, represent a bet that growth can come from a broader portfolio. Whether that bet will lift the stock above its current ceiling depends on how quickly investors see the next leg of the story — and whether it proves wider than the GPU boom that got Nvidia here.
Ad
Nvidia Stock: New Analysis - 27 May
Fresh Nvidia information released. What's the impact for investors? Our latest independent report examines recent figures and market trends.
