Nvidia (NASDAQ: NVDA) reversed course sharply on Monday. The stock climbed about 3% shortly after the opening bell, then dropped roughly 2% once the Groq news came out.

Earlier in the day, Nvidia announced that its Groq 3 LPX rack is now being produced at scale, turning technology from the largest acquisition in the company's history into customer-ready hardware. Nvidia spent $20 billion last December to buy assets from AI chip startup Groq. The racks are headed to neocloud company Nebius Group (NASDAQ: NBIS), where they will operate alongside Vera CPUs and Rubin GPUs. Nebius is one of the newer "neocloud" providers — companies that rent out AI computing capacity and have become significant buyers of high-end AI chips.

Nvidia senior director Dion Harris told reporters the systems should be online later this year. Nvidia is pursuing faster inference, an area where speed matters for AI agents — especially coding systems that need to respond without long delays. Inference — running a trained model to generate its output, rather than the training process that creates it — has grown into the dominant form of AI computing as companies move models from development into everyday use. The company says cloud providers can charge more for tokens that come with stricter speed and latency requirements. Tokens — the pieces of text that AI models read and produce — are the basic billing unit for AI cloud services.

Nvidia is now worth close to $5.2 trillion, based on FactSet (NYSE: FDS) data, making it the biggest company in the world by market value. The shares had gained nearly 7% in August through Friday's close, while the S&P 500 had risen around 2.5%. Data from London Stock Exchange Group (LSE: LSEG) show analysts expect quarterly revenue and earnings to reach about twice their levels from the same period last year.

Nvidia puts Groq 3 into customer systems as faster AI output creates another paid service

Each Groq 3 LPX rack holds 256 individual Groq 3 chips. Nvidia says the full setup can generate 3,400 tokens per second, based on Artificial Analysis benchmarks. Each processor has 500 megabytes of SRAM directly on the chip, keeping data close to the processor and cutting slowdowns from slower memory. SRAM is much faster than the high-bandwidth memory used on GPUs but also far more expensive per unit of capacity, a trade-off Groq built its original chip architecture around.

Samsung Electronics (KRX: 005930) manufactures Groq's chips, while Taiwan Semiconductor Manufacturing (NYSE: TSM) produces Nvidia's GPUs.

During the call, Harris explained how Nvidia believes cloud providers can make money from that speed. "For folks who are serving tokens, it unlocks the ability to offer premium tiers of service for those users and those customers who actually demand the most latency-sensitive" service agreements, he said. Cloud companies can then create an additional price level for customers who want quicker AI responses instead of standard service speeds.

Nvidia is also shipping out more of its Vera Rubin systems, which entered production earlier this year. When Vera Rubin and Groq 3 LPX were shown in March, CEO Jensen Huang said Blackwell and Vera Rubin together could generate $1 trillion in total sales through 2027. Huang also said he planned to give 25% of the data-center space used for coding workloads to Groq processors. "The rest of my data center is all 100% Vera Rubin," Huang said.

Wall Street raises the bar for Nvidia while analysts stick with high price expectations

Rosenblatt Securities kept its Buy rating on Nvidia and left its share-price target at $325.00. Nvidia was trading at $214.72, while InvestingPro put its fair-value estimate at $259.96.

Rosenblatt expects Nvidia's fiscal second-quarter 2027 report to come in above Wall Street expectations for both revenue and earnings, and it also expects the earnings release to push the shares higher.

Wedbush analyst Matt Bryson, who has an Outperform rating on Nvidia, expects the stock's recent rally to keep going after Wednesday's results. Bryson also believes Nvidia will come in above its own guidance and give an October-quarter outlook that beats what Wall Street currently expects.

"We again expect NVDA to exceed its guidance and to guide October above Street, given 1) a strong hyperscale spending backdrop that firmed further through CQ2 earnings, and 2) a supply position we continue to view as the best in the industry at a point where component and material access, not end demand, is defining shipments," Bryson wrote.

Nvidia is due to report its fiscal second-quarter numbers Wednesday after the market closes, and Cryptopolitan will be reporting the results live.