AI chip champion Nvidia (NASDAQ:NVDA) delivered another blockbuster quarter, but the more important conclusion for UK retail investors is not simply that the chip designer beat forecasts. It is that Nvidia is increasingly becoming the infrastructure platform for the entire AI economy, from GPUs and networking to inference, CPUs, software and AI-factory infrastructure. For now, the numbers favour the bulls.
The results released after the US market close on 26 August showed revenue of $96.2bn, up 106% year on year, while adjusted EPS of $2.22 beat the roughly $2.09 consensus. Data Centre revenue rose 117% to $89bn.
| Nvidia (NASDAQ:NVDA) | Price: $222.32 (~+6% after-hours) | Market cap: ~$5.37tn |
The bigger surprise was the outlook: Nvidia expects $108bn of Q3 revenue, comfortably ahead of the roughly $104bn consensus before the results. The shares initially fell after the release, partly because Q3 gross-margin guidance of 74% was below the 75% achieved in Q2, before rebounding as investors absorbed the strength of the revenue outlook and management commentary. Reuters reported a roughly 4.2% after-hours gain at one point.
The numbers that matter
| $bn except EPS | Q2 FY26 | Q1 FY27 | Q2 FY27 | YoY |
| Revenue | 46.7 | 81.6 | 96.2 | +106% |
| Data Centre | 41.1 | 75.2 | 89.0 | +117% |
| Gross margin | 72.4% | 74.9% | 75.0% | +2.6pp |
| Adjusted EPS | 1.01 | 1.87 | 2.22 | +120% |
| Q3 revenue guidance | — | — | 108.0 | +89% YoY |
Source: Nvidia; consensus estimates from market data providers.
The most striking number is arguably $89bn of Data Centre revenue. This is no longer primarily a graphics-chip story. Nvidia is supplying the computing infrastructure required by hyperscalers, frontier AI laboratories, enterprises, sovereign AI projects and increasingly specialised AI companies.
Why did the shares initially fall?
This is an important lesson for retail investors.
Nvidia did not report a disappointing quarter. It reported numbers that were substantially better than consensus. Yet the shares initially slipped.
Why? The market has moved beyond asking whether Nvidia can beat expectations. It is asking whether Nvidia can beat extremely high expectations by enough to justify the valuation.
The Q3 gross-margin forecast of 74% was one initial concern. Nvidia also faces sharply rising memory costs, while investors continue to debate the sustainability of enormous AI infrastructure spending.
But the earnings call changed the tone.
Nvidia subsequently pointed to accelerating demand, the Vera Rubin product cycle and a much larger opportunity in AI inference and agents. The shares moved higher in extended trading, and Nvidia was up more than 6% in Thursday trading according to Reuters.
That reversal matters: investors appear increasingly willing to look through modest near-term margin pressure if revenue growth remains exceptional.
Jensen Huang’s central message: AI is becoming an economy
CEO Jensen Huang’s most important statement was probably:
‘AI has reached its inflection point. It’s doing useful work. Its tokens are productive and profitable. Now, compute is revenue.’
That encapsulates Nvidia’s investment thesis.
The first phase of generative AI involved companies experimenting with models. The next phase is about AI inference, agents and commercially productive workloads.
Huang argued that AI agents can require dramatically more compute because they reason, plan and use tools through multiple interactions. During the earnings call he suggested the compute requirement for an agent can be 15 to 100-times that of a human using a comparable application, depending on the task.
If that proves correct, AI inference could become a second enormous demand wave after model training.
That is potentially more important for Nvidia than another quarter of 100%-plus revenue growth.
Nvidia is becoming much more than a GPU company
The company is deliberately moving up the AI value chain.
Its ecosystem now encompasses:
- Compute: Blackwell and Vera Rubin GPUs, plus Vera CPUs
- Networking: Spectrum and high-speed interconnects
- Inference: Groq 3 LPX and other inference technology
- Software: CUDA, CUDA-X, AI Enterprise and agent tooling
- Physical AI: robotics, autonomous vehicles and industrial AI
Infrastructure: complete ‘AI factory’ systems and design platforms
Financing: partnerships designed to mobilise capital for AI infrastructure
Nvidia says Vera Rubin is already entering full production, with systems running at customers including CoreWeave, Google Cloud, Microsoft Azure, Oracle Cloud Infrastructure and Nebius.
This diversification is strategically important because hyperscalers, such as Microsoft, Amazon and Alphabet are developing their own silicon.
Nvidia’s response is effectively: don’t compete with us at the chip level — use our entire platform.
Huang described Nvidia as an ‘entire AI factory platform’ spanning the AI lifecycle and said he has ‘100% confidence’ customers will continue using Nvidia compute for a long time.
The $500bn question
One of the most interesting issues raised during the earnings call was Nvidia’s growing involvement in financing the AI ecosystem.
The company has announced partnerships involving Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR designed to mobilise more than $500bn of third-party capital for AI infrastructure over time.
This addresses a genuine problem: AI infrastructure requires enormous amounts of capital.
But it also creates an obvious investor question.
Is Nvidia financing the ecosystem because demand is organically enormous — or does some of the apparent demand depend on Nvidia helping customers obtain the money to buy Nvidia’s products?
BofA analyst Vivek Arya raised precisely this issue, highlighting the potential risks from Nvidia’s commitments and what has become known as ‘circular financing’. Before the results, Arya warned that a beat-and-raise might not be enough because investors were increasingly focused on earnings quality and the company’s financial commitments.
Management’s response is that these arrangements are primarily about accelerating infrastructure deployment and securing supply capacity rather than artificially manufacturing demand.
That distinction will matter increasingly as Nvidia gets larger.
The other constraint: memory and margins
Nvidia’s gross margin remains extraordinary at 75%, but investors should not assume that level is permanent.
Q3 guidance is 74% ±0.5 percentage points, while higher memory costs are becoming an increasingly important industry issue. Analysts have also been looking at whether custom silicon from hyperscalers can eventually pressure Nvidia’s pricing power.
The important point is that a 1–2 percentage-point margin reduction is not necessarily a bearish signal if revenue is growing much faster than expected.
The risk comes if margin compression combines with a slowdown in Data Centre growth.
That would create a double hit to earnings expectations.
The biggest unanswered question: how long can this growth last?
Nvidia gave investors an unusually bullish long-range signal: management indicated that fiscal 2028 revenue could grow by around 70%.
That compares with analyst expectations of roughly 45%, according to reports following the results. Goldman Sachs subsequently raised its price target from $285 to $300, citing a clearer path to outperformance. Citi raised its target to $315.
That is a major statement because Nvidia is already a roughly $5tn company.
The debate has therefore shifted from:
‘Can Nvidia grow?’
to:
‘Can a company of this size continue growing at anything approaching this rate?’
So far, the evidence remains remarkably favourable.
Forward valuation
At around $220–225 after the post-results rally, Nvidia’s market capitalisation is roughly $5.3tn. Reuters noted that the shares trade at around 18x forward earnings, while other market-data measures put forward PE closer to the low-20s depending on the earnings period and methodology.
| Measure | Nvidia |
| Market value | ~$5.1tn |
| Q2 revenue growth | 106% |
| Q2 Data Centre growth | 117% |
| Q3 revenue guidance | $108bn |
| Q3 consensus before results | ~$104bn |
| Forward PE | roughly 18–24x, depending on estimate |
| Goldman Sachs target | $300 |
| Citi target | $315 |
| BofA target | $350 |
| Cantor target | $350 |
That valuation looks less demanding when compared with Nvidia’s growth than the headline PE suggests.
For example, if earnings continue compounding rapidly, a 20-ish forward earnings multiple can be quite reasonable for a company still growing profits at extremely high rates.
But there is an important catch.
The valuation only works if the earnings denominator keeps rising.
If Nvidia’s earnings growth eventually slows towards 20%–30%, a 20-plus multiple might become harder to defend.
What analysts are saying
The analyst response has been overwhelmingly positive.
Goldman Sachs said the results provide a ‘clearer path’ for Nvidia to outperform the semiconductor sector, while raising its target to $300.
BofA remains particularly bullish, with a $350 target, although Arya continues to highlight the balance-sheet and financing questions.
Cantor’s CJ Muse has gone even further, arguing that Nvidia is being valued too conservatively relative to its growth opportunity. Before the results he described the shares as trading at roughly 14 times projected 2027 earnings and 10 times 2028 expectations, with a $350 target.
That contrasts with the more cautious interpretation: Nvidia may be cheap only if the extraordinary growth trajectory survives the enormous scale of the business.
Nvidia’s place in the AI ecosystem
The investment case is increasingly easier to understand as a chain:
AI models → AI agents → inference → more compute → more data centres → more networking → more memory → more power → more Nvidia systems
Nvidia sits close to the centre of almost every stage.
That gives investors exposure not only to AI model developers but also to the infrastructure spending required to run them.
It also explains why Nvidia’s earnings matter to investors in companies, such as Microsoft, Amazon, Alphabet, Meta, Broadcom, TSMC, SK Hynix and other AI infrastructure suppliers.
The crucial distinction is that Nvidia is increasingly selling the whole system, rather than simply the accelerator.
Bull case vs bear case
| 🐂 Bull case | 🐻 Bear case |
| AI inference and agents create a second demand wave | AI capex eventually produces disappointing returns |
| Vera Rubin drives another major upgrade cycle | Hyperscalers increasingly use custom chips |
| Demand remains well above Nvidia’s supply | Memory and infrastructure costs squeeze margins |
| 70% FY2028 growth proves conservative | 70% growth proves unsustainable at $5tn scale |
| CUDA/platform ecosystem maintains Nvidia’s moat | Open models and alternative accelerators weaken the moat |
| Broader customer base reduces hyperscaler concentration | Financing commitments increase balance-sheet risk |
| Buybacks support EPS growth | Valuation contracts if growth expectations fall |
What UK retail investors should watch next
For investors holding Nvidia through a Stocks & Shares ISA, SIPP or taxable account, the next few quarters are likely to revolve around five numbers rather than the headline EPS beat:
1. Data Centre growth: can 117% growth remain above 100%?
2. Rubin: how quickly does the new architecture replace or complement Blackwell?
3. Gross margin: does the business stabilise around the low-to-mid 70s?
4. AI inference: does inference become a genuinely larger source of demand rather than simply a management narrative?
5. Customer economics: are Microsoft’s, Meta’s, Google’s, Amazon’s and AI-lab investments generating enough revenue to justify continued capex?
Investor verdict
The Q2 results strengthen the Nvidia bull case — but they do not eliminate the biggest risk.
Operationally, this was an exceptional quarter. Revenue exceeded $96bn, Data Centre sales approached $90bn, margins remained around 75% and Q3 guidance points to another record quarter. Most importantly, management believes demand is still substantially ahead of supply.
The bigger investment thesis is that Nvidia is evolving from the leading AI chip company into the infrastructure platform underpinning the AI economy.
That is potentially much more valuable.
But at roughly $5.3tn, investors should remember that Nvidia no longer needs merely to execute well. It needs to continually exceed an extraordinary growth trajectory that is already embedded in the share price.
For long-term investors, the Q2 report is therefore encouraging rather than definitive.
Bull case: AI becomes a new computing platform and Nvidia captures an unusually large share of the resulting infrastructure spending.
Bear case: AI investment eventually normalises, custom silicon takes share, margins fall and Nvidia’s valuation multiple contracts simultaneously.
For now, the numbers favour the bulls. The key question is no longer whether the AI boom is real. It is whether Nvidia can remain the primary financial beneficiary as that boom moves from experimentation into mass commercial deployment.
Disclaimer: The author Steven Frazer has a personal interest in Nvidia.
You might also like:
