New York, May 19, 2026, 13:33 (EDT)
- Nvidia edged higher before Wednesday’s results, while AI cloud name CoreWeave fell after Google and Blackstone moved into its market.
- The chip trade firmed, but the rally is now being tested by valuation, bond yields and the pace of data-center construction.
- Google’s new TPU cloud venture puts more attention on inference, the work of running AI models after they have been trained.
Artificial intelligence stocks were split Tuesday afternoon, with Nvidia up modestly before results investors see as a check on the whole AI trade, while CoreWeave fell after Alphabet’s Google and Blackstone laid out a new AI cloud venture. U.S. trading was in regular session; the NYSE’s core session runs from 9:30 a.m. to 4 p.m. Eastern time, and May 19 is not listed as a 2026 exchange holiday.
The matter is simple and not small: Nvidia has become the main public-market proxy for spending on artificial intelligence, software and chips used to perform tasks that once needed human judgment. LSEG data cited by Reuters showed analysts expect Nvidia’s April-quarter revenue to rise 79%, with adjusted profit up 81.8%, as Microsoft, Meta and other large cloud buyers keep spending on data centers.
That spending has supported chip shares for months. By early afternoon, Nvidia was up about 0.8%, AMD gained 0.5% and the iShares Semiconductor ETF rose 1.8%; Broadcom fell 1.2%, Microsoft lost 1.3% and Palantir slipped 0.5%. Micron, tied to memory chips used in AI servers, jumped nearly 5.9%.
The sharper move came in AI cloud. CoreWeave, a provider of rented computing power used to train and run AI models, dropped 3.7% after Google and Blackstone announced a U.S.-based venture that will offer data-center capacity and Google’s Tensor Processing Units, or TPUs, custom chips built for AI work, through a compute-as-a-service model, meaning customers rent processing power rather than buying the hardware.
Blackstone said it would commit an initial $5 billion in equity and expects the venture to bring 500 megawatts of capacity online in 2027. Blackstone President Jon Gray called it “a generational opportunity,” while Google Cloud CEO Thomas Kurian said it “helps meet growing demand for TPUs.” Blackstone
The Google-Blackstone plan adds a harder competitive question for Nvidia. Its core position in training AI models remains large, but investors are now watching inference — the running of already-trained AI systems — where cost matters more and where AMD, Google and Amazon are trying to win more work. Gabelli Funds portfolio manager John Belton told Reuters the question is whether “the Nvidia ecosystem” stays “as dominant” as new inference workloads spread. Reuters
There is also a physical constraint. Elazar Advisors analyst Chaim Siegel told Reuters some customers want more graphics processors, or GPUs, the chips Nvidia sells for heavy AI workloads, but “don’t really have the data centers” ready to house them. That matters because chip orders can slow if buildings, power connections or cooling systems lag. Reuters
The software side was less one-way. Reuters reported that several U.S. software stocks rose earlier Tuesday as investors reassessed which firms are most exposed to AI disruption and which may benefit from new products or productivity gains. Mizuho’s Gregg Moskowitz said he still saw “very attractive investments in software” for patient investors. Reuters
ServiceNow’s early move faded, however, and the stock was down 2.4% by early afternoon, while Salesforce slipped 0.4% and Workday fell 0.6%. That left the market with a familiar split: AI infrastructure still attracts money, but investors are more selective when the story moves to software margins and customer demand.
But the risk is that the numbers are already too good. A Reuters markets column noted that Nvidia is due to report after Wednesday’s close with a market value around $5.5 trillion and expectations near an 80% revenue rise; it also said the shares have fallen immediately after the last three earnings reports despite strong results. Ameriprise strategist Anthony Saglimbene wrote that higher yields could “challenge what investors are willing to pay” for long-duration growth stocks, whose value rests heavily on future earnings. Reuters
That is the downside case for AI stocks into the report. Nvidia could beat forecasts and still fail to lift the group if investors focus on slower data-center buildouts, higher borrowing costs for cloud customers, rising memory and packaging costs, or tougher competition from custom chips.
For now, the market is not abandoning the trade. It is narrowing it. Nvidia’s result will show whether that is discipline, or the first crack in a crowded bet.
