AI Computing Capacity Set for Major Expansion as Data Centers Scale

What Happened

A significant expansion of artificial intelligence computing infrastructure is underway globally, with new data centers and chip deployments set to bring online levels of processing power that represent a substantial leap over current capacity, according to reporting by The New York Times. The scale of investment arriving in 2025 and 2026 is enabling AI systems to reach milestones that earlier hardware generations could not support.

The New York Times reports that in 2023 an AI system passed the bar exam, and that by 2025 the technology had progressed to assisting scientists in ways that mark a further tier of capability. The report attributes this progression directly to the volume of compute becoming available rather than solely to algorithmic improvements.

Background

The AI hardware buildout has been driven by a sustained surge in capital investment from technology companies, cloud providers, and governments over the past two years. Nvidia has remained the dominant supplier of graphics processing units used for AI training and inference workloads. The company's position as the primary hardware provider for large-scale AI has made it central to the financing of the broader AI ecosystem.

Investor Mark Cuban, commenting separately in a Benzinga report, compared Nvidia's current role to infrastructure financing during the dot-com era, describing the company as the entity "funding everyone and anyone" pursuing AI development. Cuban noted that a rival chip breakthrough could disrupt that position, though he did not name a specific competitor or timeline.

Separately, sharp declines in the share prices of several chip makers have prompted investor concern about whether market valuations in the AI sector remain sustainable. The BBC reported that the falls have stoked questions about whether investor enthusiasm for AI-related companies is fading, though the declines have not halted the underlying infrastructure buildout.

What the Numbers Show

The New York Times report frames the current moment as one defined by the sheer volume of computing power arriving simultaneously rather than by any single product release. Data center construction timelines and chip delivery schedules from major suppliers indicate that the bulk of capacity ordered during the 2024 and 2025 investment cycle is becoming operational across this period.

The scale of capital committed to this infrastructure is reflected in company valuations. China's Moonshot AI, developer of the Kimi large language model series, reached a reported valuation of 35 billion dollars following its latest funding round, according to Moneycontrol. The funding came after the company released its Kimi K3 model and surpassed its original fundraising target, illustrating continued private market appetite for AI investment despite public market volatility in chip stocks.

What It Means in Practice

The increase in available compute affects both training and deployment of AI models. Larger training runs allow developers to build systems with capabilities not achievable at smaller scales. Increased inference capacity allows those systems to be deployed to more users simultaneously and to handle more computationally intensive tasks in real time.

The New York Times report notes that AI systems have already been applied to scientific research tasks in 2025, a category of application that requires sustained, high-volume computation and that was not practically feasible for most organizations at the hardware availability levels of two years ago.

WeRide, the autonomous vehicle company listed on Nasdaq, also announced in July 2026 a physical AI model it calls WITT, or World Intelligence Toward Truth, designed to process real-world sensor data for autonomous driving applications. The announcement illustrates how expanded compute availability is enabling AI deployment beyond text and image tasks into physical and industrial environments.

What Comes Next

Further data center capacity from major cloud providers is scheduled to come online through the remainder of 2026, with chip suppliers including Nvidia having previously announced next-generation hardware release timelines that fall within that window.