Nvidia Q4: Slowing Growth vs AMD’s Power Move

While American/Taiwanese chipmaker Nvidia Corporation (NVDA) showed overall strong results for its Fiscal Year (FY) 2026 in its release after markets closed on the 25th of February 2026, stock performance immediately afterwards was relatively modest for two reasons. Firstly, the company largely performed as expected. Second, the gap between its rival – AMD – seems to be narrowing.

Trend Analysis

In FY 2026, the company continues to grow in dependence on revenues from the AI boom, which it services through the “Compute & Networking” segment.

Websim is the retail division of Intermonte, the primary intermediary of the Italian stock exchange for institutional investors. Leverage Shares often features in its speculative analysis based on macros/fundamentals. However, the information is published in Italian. To provide better information for our non-Italian investors, we bring to you a quick translation of the analysis they present to Italian retail investors. To ensure rapid delivery, text in the charts will not be translated. The views expressed here are of Websim. Leverage Shares in no way endorses these views. If you are unsure about the suitability of an investment, please seek financial advice. View the original at

Source: Company Information; Leverage Shares analysis

Multiple years of focus have yielded very strong operating income contribution from the AI sector, which comfortably exceeds its revenue share. Meanwhile, the “Graphics” segment – which caters to consumers, gamers, and graphics designers et al – is a shadow of its self barely six years ago. If current trends continue, FY 2027 could see this segment run at 5-6% of revenue contribution and below 5% of operating income contribution.

The massive reliance on corporate spends in AI lends itself to an emerging pattern of seasonality: in the third quarter of FY 2026, net income per share was trending at closing out the FY with an 8% growth. With sales in its final quarter, the year closed with a net 65% growth over the previous FY.

Source: Company Information; Leverage Shares analysis

One development stands out in the long-term trend: FY 2024 marked a headlong and variegated rush into the AI space, which massively boosted the company’s net income growth well above that of revenue growth on the back of strong demand attracting premium prices. The gap between net income growth vis-à-vis revenue growth narrowed in FY 2025 and has nearly flattened in FY 2026.

Two factors might lie behind this trend: firstly, the number of hyperscalers and LLM (“Large Language Model”) developers have cooled off over the past couple of years to form a set number of allied firms making consolidated buys via a massive web of “circular AI deals” largely led by Nvidia itself¹, which doesn’t quite help the company in commanding a premium on products.

Second, the company seems to have gone from being an industry disruptor – which essentially kicked off consideration of “AI Compute” as a product and segment in the market that is distinct from regular consumer products – to an industry standard. In FY 2023, R&D (“research and development”) accounted for 27% of revenue in a year with 0% YoY (“Year-on-Year”) growth. In FY 2026, R&D is 9% of revenue. Stock-based compensation, meanwhile, exhibits steady 30-35% growth over the last three FYs and constitutes only 3% of revenue in FY 2026 – the lowest level since FY 2019. Meanwhile, net income share remains consistent at 56% across both FY 2025 and FY 2026.

The company’s revenues are overwhelmingly defined by sales to U.S.-headquartered clients now.

Source: Company Information; Leverage Shares analysis

When considering China as well as Taiwan together – and considering the possibility of product flows between the two – Nvidia’s revenue share remains locked at 29% in both FY 2025 and FY 2026. On the other hand, sales to all clients headquartered in other countries – and the budding AI infrastructure therein – has dropped to near-insignificance. Now, more so than over, Nvidia is resolutely an American supplier to American companies.

There are potential downsides to being an “industry standard” in such circumstances, as the recent Meta-AMD deal signifies.

How AMD Closes the Gap and Why It Could

On the 24th of February, AMD and Meta Platforms announced a deal² wherein Meta will be purchasing 6 gigawatts (GW) of its Instinct Graphics Processing Units (GPUs) for up to $60 billion over the next five years. AMD sweetened the deal for Meta by taking a page out of Nvidia’s playbook over the past couple of years and turbocharging it: it offered Meta the right (but not the obligation) to buy up to 10% of itself – 160 million shares – at a strike price of $0.01, contingent on deployment milestones.

At the centre of the deal is AMD’s Instinct MI450 GPU, which is being positioned as a rival to Nvidia’s recently unveiled industry-specific Rubin platform, as well as 6th Gen EPYC “Venice” CPUs. The MI450 is the cornerstone of AMD’s Helios rack architecture, designed to compete directly with Nvidia’s NVL72 (“Rubin”-based) systems.

It bears noting that the MI450 doesn’t have strong advantages versus the “Rubin” in all parameters:

Source: Leverage Shares analysis

While Nvidia’s Rubin Ultra offers higher raw performance, the MI450 holds a significant lead in next-generation industry standard High Bandwidth Memory (HBM4) capacity and raw bandwidth. This is a deliberate play for Meta’s Llama 5 and Llama 6 training requirements, which are increasingly memory-bound rather than compute-bound. The positioning AMD is wagering a substantial portion of its company on is that Meta’s models will run more efficiently on the larger memory footprint.

By delineating model training versus model inference, AMD is working on splitting the market and offering price-matched products that Nvidia might have difficulties in rationalizing its product lineup’s pricing for. The Meta deal effectively implies 3 million GPUs which, when considering the contract “floor” of $60 billion, translates to around $20,000 per GPU. While integrating the rack architecture – which includes the “Venice” CPU and Pensando – the “blended” unit price shifts to the $22,000-28,000 range. If Meta were to exercise its warrants at AMD’s current stock price and then sell the shares into the market, this would yield a “blended” price in the $7,000-8,000 range.

In contrast, the “blended” price of the Rubin, NVLink 6 and Vera CPU – using similar methods – yields around a unit price in the $45,000-55,000 range. Even without the warrant discount, AMD is thus offering Meta a massive discount to clients in order to prove its product proposition.

Historically, Nvidia’s CUDA parallel computing platform and programming model was the “moat” that had companies locking on to and selecting by default the company’s next set of products. Meta’s decision to optimize their stack for AMD’s open-source ROCm platform makes it significantly easier for other companies to switch to AMD, which helps fill in the “moat” and potentially end one factor that carried forward Nvidia’s market dominance.

What works for Meta’s needs might also work for the likes of Google, Microsoft and Amazon, thus making the battle a far more complex proposition than just supplying the market with increasingly higher-priced products across the board.

What Lies Ahead

The battle for the “AI compute” market – and the split being attempted/engineered by AMD’s aggressive new manoeuvres in being “good enough” at a reasonable price for a specific segment – is a marathon and not a sprint. By no means is Nvidia currently suffering; on the contrary, inventory trends indicate a six-year high:

Source: Company Information; Leverage Shares analysis

The near-perfect split between raw material, work in-process and finished goods seen in FY 2025 is now slightly less orderly but the company’s high connectivity with its client base indicates that the 112% growth in inventory is highly justifiable and likely to be sold out. For its Q1 2027 outlook (currently ongoing), the company forecasts a 15% increase in quarterly revenue to around $78 billion. Stock-based compensation is forecasted to be around 30% that of FY 2026 at $1.9 billion and – in a possible sign that Nvidia is committing to doing a phase-wise pullout from China – will not be assuming “any Data Center compute revenue from China”.

In the inference market at least, Nvidia’s Rubin promises to deliver up to a 10x reduction in inference token cost while being comparable with the Blackwell platform. This should mark a strong inducement for hyperscalers and LLM companies to buy in. But over the long run – as the revenue vs net income gap analysis indicated – the splitting of the market is poised to continue to enforce long-term transformation on the company’s bottom line.

More so than a “growth” stock with high and volatile investor conviction, this might be the beginning of a phase wherein it settles into “value” stock territory – which invariably will mean lower but stable conviction and trajectory. The lack of a significant post-release price bump isn’t necessarily a lack of conviction; it’s a recognition of value proposition. More so than before, Nvidia depends on what a set handful of clients decide on spending their money on – rather than being a compelling purchase decision that attracts customers from near and far.

Meanwhile, there is plenty to suggest that any potential for a massive drop will be smoothed by the company itself: in FY 2026, the company expended over $41 billion via its stock repurchase program, which helped maintain shareholder value. It still has over $58 billion left over from the previous authorization along and over $96 billion in free cash flow while the board is seemingly content in continuing to pay $0.01 (i.e., the same price as AMD’s stock warrants to Meta) in quarterly dividends to investors.

Professional investors in Europe might consider the +3x Long NVIDIA ETP (NVD3) and the -3x Short NVIDIA ETP (NV3S) during bullish and bearish trends in NVDA’s price. To potentially capitalize on major tech stocks seemingly driving the market currently, the 5x Long Magnificent 7 ETP (MAG7) and the -3x Short Magnificent 7 ETP (MAGS) are at hand.

Furthermore, the NVIDIA Options ETP (LSE ticker: NVDI) seeks to generate monthly income by buying NVIDIA shares and selling put options on them. Also available is the Magnificent 7 Options ETP (MAGO), which invests in each Magnificent 7 constituent’s respective Options ETP in an equally-weighted manner.

Footnotes:

“Meta and Google: Revenue Growth Concerns, Costly AI”, Leverage Shares, 3 November 2025
“Meta and AMD Partner for Longterm AI Infrastructure Agreement”, Meta Newsroom, 24 February 2026

Your capital is at risk if you invest. You could lose all your investment. Please see the full risk warning here.