AI stocks are diverging sharply as investors reward strong infrastructure demand while scrutinising growth, margins and valuations across the sector. CoreWeave, Super Micro, Cerebras and Nvidia highlight an increasingly selective AI trade.
Sharp moves across CoreWeave, Super Micro, Cerebras and Nvidia are exposing a widening divide in the AI trade as investors demand stronger evidence that record infrastructure spending can translate into sustainable growth.
Artificial intelligence stocks are producing some of the largest moves in global equity markets this week, however the rally is becoming increasingly selective as investors scrutinise which companies are converting the AI spending boom into revenue, margins and long-term contracts.
CoreWeave and Super Micro Computer surged on Wednesday 12 August 2026, after both companies delivered results and forecasts that reinforced expectations of sustained demand for AI computing infrastructure. Nvidia also climbed as enthusiasm spread across semiconductor and data-centre stocks, helping lift the Nasdaq and S&P 500.
At the other end of the trade, Cerebras Systems came under heavy pressure in early Thursday trading indications after mixed quarterly results raised fresh questions over whether rapid growth in parts of its business was enough to meet the exceptionally high expectations now attached to AI-related companies.
The contrasting reactions point to an important development for equity markets. Investors are still willing to pay for exposure to the expansion of artificial intelligence, but increasingly appear to be distinguishing between companies that can demonstrate commercial demand and those where the investment case remains dependent on future growth.
CoreWeave surge shows strength of AI computing demand
CoreWeave shares jumped more than 19% on Wednesday after the specialist AI cloud provider reported second-quarter revenue of $2.58 billion, more than double the level recorded a year earlier and slightly ahead of market expectations.
Perhaps more significant than the quarterly number was the company’s rapidly expanding order book. CoreWeave ended the quarter with a revenue backlog of $104.2 billion, up from $99.4 billion three months earlier, whilst securing more than $25 billion in additional customer commitments early in the current quarter. Demand is sufficiently strong that CoreWeave says its near-term computing capacity is effectively sold out, allowing the company to negotiate new contracts on increasingly favourable terms.
The company has consequently raised its expected 2026 capital expenditure to between $35 billion and $39 billion, up from an earlier $31 billion to $35 billion range. The scale of that investment illustrates how much physical infrastructure is now required to meet demand for AI models, cloud computing and data processing.
CoreWeave’s close relationship with Nvidia is central to that expansion. Its data centres provide customers with access to Nvidia processors, making the company one of the clearest publicly traded proxies for demand further down the AI infrastructure chain.
Super Micro adds another signal from the data-centre buildout
Super Micro Computer provided further evidence that the infrastructure cycle remains strong. The server manufacturer forecast fiscal 2027 revenue of between $65 billion and $72 billion, substantially above the approximately $52.5 billion analysts had been expecting. Its quarterly gross margin also reached 17.5%, while revenue of $11.12 billion almost doubled from a year earlier despite falling slightly short of analysts’ forecasts.
The market focused heavily on the outlook rather than the revenue miss. Super Micro shares ultimately climbed around 19% on Wednesday as investors responded to expectations that spending on AI-optimised servers and data centres will remain elevated.
Other companies connected to the infrastructure buildout participated in the rally. Nebius, Applied Digital, IREN and Dell all advanced, while Nvidia rose about 3%. The Philadelphia semiconductor index gained around 2.5%.
That breadth suggests investors were not simply reacting to one company’s earnings. The results strengthened the broader argument that demand for the computing power behind generative AI remains substantial.
Cerebras highlights the other side of the AI trade
Cerebras Systems produced a very different reaction on Thursday. The Nvidia challenger reported rapid growth in its cloud operation, where quarterly revenue roughly quadrupled from a year earlier to $126 million. However, hardware sales, including its AI processors, declined to $54.1 million from $70.3 million.
Adjusted gross margin also fell to 40.6% from 46.5% in the previous quarter, while overall revenue missed market expectations despite the company raising its annual outlook.
Cerebras had risen around 41% from its $185 initial public offering price before the results, leaving a high bar for the newly listed company to clear. Its performance is being closely followed because the company has positioned its processors as an alternative to Nvidia’s dominant AI hardware. The response illustrates the increasingly unforgiving nature of AI valuations. Rapid growth alone may no longer be sufficient when investors have already priced substantial future expansion into a company’s shares.
Cisco delivered another example after forecasting continued AI infrastructure growth but seeing its shares fall in extended trading. The networking group generated $9.3 billion of AI infrastructure orders from hyperscale customers during its 2026 financial year, yet expectations surrounding the sector have become high enough that strong numbers do not necessarily guarantee a positive share-price reaction.
More than $740 billion is flowing into AI investment
The scale behind these market moves is considerable. Big technology companies are expected to spend more than $740 billion this year as they expand data centres, processors, networking equipment and other infrastructure required to develop and operate artificial intelligence systems.
This spending is creating a broad ecosystem of potential beneficiaries extending far beyond the companies developing AI models themselves.
Chipmakers such as Nvidia supply the processing power. Companies including CoreWeave provide access to that computing capacity. Super Micro supplies AI servers, while networking companies such as Cisco connect increasingly complex data centres. Power, cooling, memory and data-centre operators form further layers of the same investment cycle.
The relationship between those businesses also means developments at one company can quickly affect valuations elsewhere. Strong demand reported by an infrastructure provider can reinforce expectations for Nvidia chip sales, while slowing orders or weaker margins can raise questions about the economics of the wider AI buildout.
Nvidia remains the key test for the AI market
Nvidia remains the central company in that chain and arguably the most important barometer for the broader AI investment cycle. Its shares gained around 3% during Wednesday’s rally as CoreWeave and Super Micro’s results provided further evidence that spending on AI computing remains strong. With a market capitalisation above $5 trillion, movements in Nvidia also have an unusually large influence on major US equity indices.
Attention will therefore turn increasingly towards Nvidia’s next quarterly results, due later in August. Beyond headline revenue growth, markets are likely to examine data-centre demand, margins and management’s expectations for the next generation of AI infrastructure.
The significance extends well beyond Nvidia shareholders. Evidence of accelerating demand could support valuations across semiconductors, servers, data centres and AI cloud providers. Any indication that spending is slowing could have the opposite effect.
What the changing AI trade means for traders
The latest earnings season suggests that simply identifying a company as an “AI stock” is becoming less useful as a way of understanding its likely market performance. Companies connected to the same structural trend are producing dramatically different share-price reactions depending on revenue growth, margins, order backlogs, capital requirements and management forecasts. That creates opportunity, but also considerable event risk for anyone trading individual AI-related shares.
The distinction is particularly relevant where stocks are traded through leveraged products such as CFDs. Double-digit movements following earnings announcements can magnify both profits and losses, while pre-market and after-hours price changes can lead to substantial gaps when normal trading resumes.
For traders following the AI sector, the next phase of the boom may therefore be less about whether artificial intelligence continues to attract capital and more about which companies can demonstrate that they are capturing enough of that spending to justify their valuations.
For now, CoreWeave and Super Micro have provided strong evidence that the infrastructure buildout remains active. Cerebras and Cisco demonstrate the other side of the equation: when expectations are already exceptionally high, merely participating in the AI boom may no longer be enough.
