Astera Labs gave investors the quarter they were waiting for, then ran into the harder problem: a stock can outrun even a real AI infrastructure story.
On August 4, Astera Labs reported second-quarter revenue of $392.4 million, up 104% from a year earlier and 27% from the prior quarter. The company had guided for $355 million to $365 million, so this wasn't a small beat. It then guided the current quarter to $540 million to $560 million, with non-GAAP earnings of $1.16 to $1.21 a share. Investor's Business Daily said Wall Street had been looking for about $417 million in revenue and $0.81 a share.
You would expect numbers like that to settle the argument. They didn't. Astera's shares had already been bid up hard into the report, including a 12.7% jump during the regular session before the earnings release, according to Investor's Business Daily. The stock then slipped after hours and kept trading choppily as investors worked out whether the beat was enough for a company already priced like one of the cleanest AI supply chain winners.
Demand isn't the obvious problem. Astera sells the connectivity parts that help AI systems move data inside and across server racks: Aries PCIe and CXL retimers, Scorpio fabric switches, Taurus signal conditioners and Leo CXL memory controllers. In its own release, the company said Q3 growth would be driven by the Scorpio X-Series 320-lane fabric switch production ramp, and Chief Executive Jitendra Mohan said Scorpio would become Astera's largest product family one quarter earlier than expected.
That's a real business signal. It also makes the stock's reaction more interesting than a simple disappointment story.
The AI trade is getting less forgiving #
Datadog gave investors a cleaner example of what happens when one piece of the AI-growth story cracks. The Wall Street Journal reported that Datadog expects annual revenue of $4.45 billion to $4.47 billion, above its prior forecast but below Wall Street's $4.69 billion estimate, because one of its biggest customers is reducing usage. The company described that customer as a leading AI company that had just signed a nine-figure renewal. Investor's Business Daily said investors were focused on whether OpenAI was behind the lower usage.
That was enough. Datadog stock fell sharply even though second-quarter revenue rose 36% to $1.12 billion and adjusted earnings beat estimates. If you're holding expensive AI-adjacent software, this is the part that should bother you: usage can slow before the headline customer relationship breaks. A renewal doesn't always mean spending keeps climbing at the old pace.
Nokia's July 23 report told a different story. Reuters reported that comparable operating profit rose 18% to 434 million euros, ahead of the 382 million euros analysts polled by LSEG expected, while comparable net sales reached 4.82 billion euros. Nokia also said sales to AI and cloud customers more than doubled to 446 million euros, and it booked 2.8 billion euros in new AI and cloud orders.
There was a catch. Reuters quoted Nokia CEO Justin Hotard saying demand remained strong while supply was the main constraint, and Benzinga, citing a Bloomberg interview, reported that Hotard expects memory shortages to continue through 2027. That isn't the same problem Datadog has. It isn't a vanishing customer. It's a supply chain that can turn demand into higher costs and longer lead times before it turns into clean revenue.
Astera's issue is valuation, not demand #
Astera doesn't fit neatly beside either of those cases. Datadog's worry is customer usage. Nokia's worry is memory supply. Astera's worry is simpler and more dangerous for anyone buying late: expectations may already have done too much work.
The company is spending like it sees more room ahead. Astera's Q2 operating expenses were $198.3 million on a GAAP basis, and it guided Q3 GAAP operating expenses to $232 million to $236 million. That jump makes sense if Scorpio, Taurus, Leo and custom connectivity designs keep winning slots in AI racks. It also gives investors something to question when the stock has already run hard. Fast growth can carry a lot. It can't carry every price.
Here's the thing: this isn't Wall Street suddenly deciding Astera Labs is weak. The company's own figures point the other way. Revenue more than doubled, margins stayed high, and the next-quarter guide was far ahead of consensus. The market reaction says investors are becoming less willing to pay any price for AI infrastructure exposure, even when the company underneath is executing.
If you're watching the AI supply chain, don't flatten these stories into one neat selloff. Datadog shows usage risk. Nokia shows component risk. Astera shows valuation risk. Only one of those came with a quarter that was better than advertised and a guide that forced estimates higher. That's the useful distinction. Astera Labs didn't give investors bad news. It gave them the kind of good news they had already paid for.
Also read: Situational Awareness Pours $400 Million Into Chip Startup Source Foundry • Unitree's Shanghai IPO Prices China's First Humanoid Robot Maker at $9 Billion • AI Model Distillation Becomes the New Battleground Between the US and China