When Nvidia’s latest quarter hit the tape, jaws dropped — again. The chipmaker’s data‑center division pulled in US$41.1 billion of revenue, up 56 % from a year earlier. That’s the kind of number that makes investors feel like they’re riding shotgun on a rocket. Yet for all the fireworks, the company actually missed analysts’ sky‑high expectations by around US$200 million. And underneath the headline number was a trend shift: while GPU and CPU sales slipped slightly quarter‑to‑quarter, networking gear sales (InfiniBand, Spectrum‑X and NVSwitch) almost doubled. Customers, it seems, are squeezing more juice out of existing clusters instead of buying endless new silicon.
Bubble talk: hype vs. hard numbers
Is this just the opening act of a much bigger show—or the beginning of a comedown? Critics point to several red flags:
- Valuations in the stratosphere. Nvidia’s market cap flirted with US$4 trillion this year, and some investors worry the stock is priced for perfection. The company guided to US$54 billion in Q3 revenue, above consensus but short of the most bullish expectations of US$60 billion. The modest beat sent the stock down 3 % after earnings.
- Grandiose spending forecasts. On the earnings call, CFO Colette Kress said cloud and enterprise customers could spend US$600 billion on AI infrastructure this year and a whopping US$3 – 4 trillion by 2030. CEO Jensen Huang doubled down the next day, essentially betting that a new industrial revolution has begun. Those numbers imply years of 50 %‑plus growth — a pace some analysts call “aspirational.” The Next Platform cautions that maintaining such explosive growth would add “several hundred billion” dollars of revenue beyond what’s realistic.
- Signs of normalisation. As the revenue base balloons, growth naturally slows. Analysts note the sequential decline in compute‑product revenue and the outsized growth in networking as evidence that customers are optimising what they have. Their more conservative scenario sees data‑center revenue reaching about US$172 billion by fiscal 2030 — huge, but nowhere near the trillion‑dollar dreams.
Yet there are also reasons to doubt a bubble is about to burst:
- Demand remains insatiable. Big Tech companies — Microsoft, Amazon, Meta and more — account for roughly half of Nvidia’s data‑center revenue. They have committed tens of billions to generative‑AI initiatives, and Huang says everything is sold out. One customer outside China even bought US$650 million worth of Nvidia’s H20 chips.
- Profitability is still high. Nvidia expects gross margins of about 73.5 %, and margins have actually edged higher despite the meteoric growth. This isn’t the early‑2000s dot‑com era when revenue growth masked massive losses.
- AI isn’t a fad. Whether training large language models or deploying generative‑AI into industry, customers view AI as strategic infrastructure. Portfolio managers told Reuters that hyperscaler capex plans suggest durable demand for Nvidia’s chips. As one analyst put it, we may simply be transitioning from mind‑blowing beats to still‑robust (but not exponential) growth.
China: a wild card at the table
The biggest question mark in Nvidia’s near‑term outlook isn’t innovation — it’s geopolitics.
The export‑control saga
- Licenses with a catch. In 2024 the U.S. government barred sales of high‑end AI chips to China, only to issue selective licenses a year later. Nvidia’s “H20” chip — a pared‑down version of its flagship accelerators — got the green light, but only if the company forks over 15 % of China sales to Washington. Nvidia’s Q3 guidance doesn’t include any H20 sales to China, and CFO Kress said the company is ready to ship US$2 – 5 billion of H20 chips if the rules are clarified.
- Potential hit to revenue. Analysts estimate that previous export restrictions cost Nvidia roughly US$8 billion in sales. The company says a resumption of H20 shipments could add US$2 – 5 billion next quarter. At the same time, local competitors like Huawei are racing to fill the void, and The Next Platform warns that lost China sales could total US$30 billion over time.
Why China matters
CEO Huang calls China the second‑largest computing market and notes that roughly half of the world’s AI researchers work there. Without access to Chinese customers, some analysts worry that Nvidia’s valuation — which already embeds sky‑high growth — may be hard to sustain. Others counter that booming U.S. and European demand could offset any drag. The truth probably lies in between: China could add several billion dollars to revenue if export restrictions ease, but even if sales remain blocked, Nvidia still has a global gold mine to tap.
The big takeaway
So, is Nvidia’s data‑center juggernaut a bubble? The answer is more nuanced than a simple yes or no. Growth is slowing, valuations are lofty, and the company’s US$3‑trillion‑plus projections seem more like stretch goals than guarantees. Yet the demand for AI infrastructure is very real, and Nvidia’s dominance — buttressed by high margins and networking innovations — is not collapsing any time soon. The biggest swing factor is China: a full return to that market could turbocharge results, while continued restrictions could accelerate local competition.
For investors, the key is to separate hype from durable fundamentals. Nvidia remains at the epicentre of the AI boom, but betting on perpetual 50 % growth and friction‑free geopolitics is a gamble. Watch the mix of compute vs. networking revenue, hyperscaler spending intentions, and how the China saga unfolds. This rollercoaster has plenty of twists left.
