Nvidia reported fiscal fourth-quarter results on Wednesday that topped Wall Street expectations, propelled by explosive growth in its data center division. Shares rose about 2% in extended trading following the announcement.

According to CNBC, the company posted adjusted earnings per share of $1.62, ahead of the $1.53 expected by analysts polled by LSEG. Revenue reached $68.13 billion, exceeding estimates of $66.21 billion and marking a 73% increase from $39.3 billion a year earlier.

The numbers underscore Nvidia’s central role in the global AI infrastructure buildout. More than 91% of the company’s total revenue now comes from its data center unit, which houses its artificial intelligence accelerators and associated networking components.

Data center revenue totaled $62.3 billion in the quarter, ahead of StreetAccount estimates of $60.69 billion and up 75% year over year. Net income nearly doubled to $43 billion, or $1.76 per share, compared with $22.1 billion, or 89 cents per share, in the same quarter last year.

Guidance signals sustained AI demand

For the fiscal first quarter, Nvidia forecast revenue of $78 billion, plus or minus 2%, well above analyst expectations of $72.6 billion. The company said its outlook does not assume data center revenue from China, signaling that growth projections are anchored in other regions amid ongoing geopolitical and export restrictions.

The guidance reinforces Nvidia’s position as the primary beneficiary of AI capital expenditures. So far in 2026, Nvidia shares are up 5%, outperforming the broader Nasdaq, which is down 0.4%. Among trillion-dollar companies, only Apple has posted gains this year, and those are modest by comparison.

Hyperscaler spending drives momentum

Investors had early visibility into AI infrastructure momentum when the four largest U.S. cloud providers — Alphabet, Amazon, Meta, and Microsoft — reported quarterly results and outlined aggressive capital expenditure plans. Based on company forecasts and analyst projections, the combined capex for 2026 could approach $700 billion as hyperscalers expand AI data centers.

In CFO commentary, Nvidia said hyperscalers remained its largest customer category, accounting for just over 50% of data center revenue. That concentration underscores both the durability of demand and the strategic importance of a handful of buyers in shaping Nvidia’s revenue trajectory.

Networking becomes a breakout growth engine

Within the data center segment, Nvidia’s networking business posted $10.98 billion in quarterly revenue, up 263% year over year. The surge reflects growing adoption of NVLink interconnect technology and Spectrum-X Ethernet switches, which enable large clusters of GPUs to operate as unified AI supercomputers. New deals with Meta contributed to the strength.

The rapid growth in networking highlights a structural shift: AI workloads increasingly depend not only on raw compute but also on high-bandwidth, low-latency interconnects. As models scale into trillions of parameters, data transfer between GPUs becomes a critical bottleneck, elevating the value of Nvidia’s integrated hardware stack.

Gaming steady, but no longer the growth driver

Nvidia’s gaming division, once its primary revenue engine, generated $3.7 billion in revenue, up 47% year over year but down 13% sequentially. Analysts have speculated that the company may delay launching a new consumer GPU this year due to memory constraints, prioritizing high-margin AI accelerators such as rack-scale systems built around its 72-GPU Grace Blackwell architecture.

Global memory shortages have emerged as a risk factor. High-bandwidth memory (HBM), essential for AI accelerators, remains supply-constrained, forcing chipmakers to allocate production toward enterprise AI demand rather than consumer graphics.

Vera Rubin on deck

Investor focus is increasingly turning to Nvidia’s next-generation rack-scale system, Vera Rubin, the successor to Grace Blackwell. CFO Colette Kress said the company shipped its first Vera Rubin samples to customers this week and remains on track for production shipments in the second half of the year.

Vera Rubin is expected to deliver 10 times more performance per watt, a critical metric as power constraints become a defining challenge for global data centers. Energy efficiency is now a competitive differentiator, as hyperscalers grapple with grid limitations and sustainability targets.

Nvidia said it is expanding manufacturing beyond Asia into the United States and Latin America to strengthen supply chain resiliency and reduce geographic concentration risk.

“These moves are expected to strengthen our supply chain, add resiliency and redundancy, and meet the growing demand for AI infrastructure,” the company said in its filing.

It added that scaling production will depend on the capacity of local manufacturing ecosystems to ramp output on time.

The shift reflects broader geopolitical pressures and export controls that have reshaped semiconductor supply chains. Nvidia’s decision to exclude China data center revenue from forward guidance further signals sensitivity to regulatory constraints.

In the automotive segment, which includes chips for autonomous vehicles and robotics, Nvidia reported $604 million in revenue, up 6% year over year but below StreetAccount expectations of $654.8 million. The modest growth contrasts sharply with the data center surge and suggests that AI-driven demand remains concentrated in cloud infrastructure rather than edge deployment.

Strategic investments and capital risk

Beyond product revenue, Nvidia disclosed that it invested $17.5 billion over the year in private companies and infrastructure funds, primarily supporting early-stage AI startups. The company acknowledged in its annual filing that those investments “may not become profitable in the near term, or at all.”

The strategy positions Nvidia not only as a hardware supplier but also as a financial backer of the broader AI ecosystem. However, it introduces balance sheet risk, particularly if venture-backed AI firms struggle to monetize at the pace implied by current infrastructure spending.

Nvidia has also taken a large stake in Intel, further entangling it in the competitive and strategic dynamics of the semiconductor industry.

A defining cycle for AI infrastructure

The quarter reinforces Nvidia’s dominance at a pivotal moment in the AI investment cycle. Hyperscaler capex remains elevated, next-generation systems promise significant efficiency gains, and networking has emerged as a high-growth adjacency.

The open question for investors is sustainability. If AI adoption continues at its current pace, Nvidia’s vertically integrated stack — from GPUs to interconnects to rack-scale systems — positions it to capture disproportionate value. If enterprise ROI slows or capital markets tighten, the scale of infrastructure commitments could come under scrutiny.

For now, Nvidia’s results suggest that AI infrastructure demand remains robust — and that the company continues to sit at the center of the global buildout.

Google is folding robotics software company Intrinsic into its main operations, a move that reflects parent company Alphabet Inc.’s push to tighten strategic focus and accelerate commercialization of AI-driven technologies.

Intrinsic began inside Alphabet’s experimental X division before spinning out in 2021 into the “Other Bets” segment, where high-risk, long-horizon projects are incubated. Its reintegration into Google marks a shift from exploratory moonshot status to a business line expected to scale within the company’s core AI and cloud ecosystem.

The economics of robotics have changed over the past decade. Industrial hardware, such as robotic arms, has become more affordable and modular, but programming complexity has remained a barrier. Configuring robots for specific manufacturing or logistics tasks often requires highly specialized engineers writing extensive custom code, limiting adoption among mid-sized manufacturers.

Intrinsic’s flagship platform, Flowstate, aims to abstract that complexity. The web-based system allows developers and operators to build robotic applications with minimal low-level coding, reducing development time and potentially broadening the pool of users who can deploy automation.

By embedding Intrinsic into Google, the platform will now leverage Gemini AI models, Google Cloud infrastructure, and collaboration with Google DeepMind. That integration opens the door to more adaptive robotics systems capable of learning from data, simulating outcomes, and optimizing workflows in real time.

The shift also points to the growing convergence between generative AI and physical systems. Large models are increasingly being positioned as orchestration layers that interpret instructions, analyze environments, and coordinate robotic actions.

Competitive Positioning in Physical AI

Alphabet’s move comes amid intensifying competition. Amazon continues expanding robotics in its fulfillment network, integrating automation deeply into warehouse operations. Tesla is advancing its humanoid robotics initiative alongside autonomous driving systems, aiming to merge AI perception with physical dexterity.

By bringing Intrinsic into Google’s core business, Alphabet strengthens its ability to offer end-to-end solutions that combine AI models, cloud services, and robotics software. This integrated stack could appeal to enterprise clients seeking scalable automation rather than bespoke engineering projects.

The November partnership between Intrinsic and Foxconn, a key supplier to Nvidia and other technology firms, to deploy AI-driven robotics for electronics assembly in U.S. facilities, signaled that the company’s ambitions are tied to production-scale manufacturing rather than proof-of-concept experimentation.

Such deployments demonstrate how AI-enabled robotics could reshape labor-intensive assembly processes, particularly in sectors under pressure to reshore manufacturing or increase supply chain resilience.

Strategic Streamlining and Capital Discipline

Alphabet has faced increasing scrutiny from investors over spending in its “Other Bets” portfolio. Folding Intrinsic into Google suggests a preference for projects with clearer monetization pathways and closer alignment with the company’s AI and cloud growth strategy.

Google’s enterprise cloud business provides a natural distribution channel. If Flowstate and related robotics tools are bundled with cloud services, Alphabet can position itself as a digital-to-physical infrastructure provider — supplying not only data processing and AI models but also the orchestration software that runs machines on factory floors.

This approach aligns with broader corporate AI adoption trends. Enterprises are moving from experimentation with generative AI toward operational integration. Robotics, particularly in logistics and manufacturing, represents one of the most tangible ways to convert AI capabilities into productivity gains.

Scaling Beyond the Lab

Intrinsic CEO Wendy Tan White said the integration would allow the company to “unlock the promise of physical AI for a much broader set of manufacturing businesses and developers.” Access to Google’s computing resources, AI research, and global enterprise relationships could significantly expand its addressable market.

The transition from moonshot to mainstream underscores a broader thesis: AI’s next competitive frontier lies not only in digital interfaces but in real-world execution. As models become more capable, the differentiator may shift to how effectively companies embed them into operational systems.

For Alphabet, bringing Intrinsic in-house signals confidence that robotics software is no longer speculative. It is being repositioned as a strategic layer within Google’s AI-first roadmap — one aimed at connecting cloud intelligence to physical infrastructure at an industrial scale.

The move illustrates how the race for AI leadership is expanding beyond search engines and chat interfaces into warehouses, assembly lines, and supply chains. There, the commercial stakes are measured not in clicks, but in throughput, efficiency, and productivity gains.

The economic payoff from artificial intelligence remains contested, with major Wall Street institutions questioning whether the technology’s rapid corporate adoption has translated into measurable gains for the U.S. economy.

Analysts at Goldman Sachs wrote that the impact of AI on U.S. gross domestic product in 2025 was “basically zero,” arguing that large language models, chatbots, and related systems did not materially contribute to the country’s officially recorded 2.2% GDP growth. The assessment, led by Goldman economist Joseph Briggs, challenges the narrative that AI is already functioning as a broad-based growth engine.

The claim lands at a sensitive moment when equity markets have poured capital into AI infrastructure providers and application-layer firms, and valuations across parts of the technology sector imply expectations of substantial productivity gains. If the measurable macroeconomic impact remains negligible, questions about a potential AI-driven market bubble are likely to intensify.

Goldman’s position highlights a distinction between capital spending and realized productivity. While technology companies are committing hundreds of billions of dollars to AI infrastructure, those outlays do not automatically translate into immediate gains in output per worker or overall GDP growth.

Other financial institutions, including Morgan Stanley and JPMorgan Chase, have expressed similar caution. Analysts at those firms have noted that much of the near-term economic benefit from AI-related investment may accrue to manufacturing economies in Asia rather than the United States.

Massive data center expansion plans from Amazon, Google, and Microsoft require advanced semiconductors, servers, cooling systems, and networking hardware. Analysts estimate that roughly three-quarters of projected Big Tech capital expenditures could directly support GDP growth in Taiwan and other Asian technology manufacturing hubs, where much of the hardware supply chain is concentrated.

This geographic distribution complicates measurement. When U.S. firms import high-value chips and components, the domestic GDP effect may be muted even if corporate revenues and market capitalizations rise.

President Donald Trump has argued that AI investments are supporting U.S. economic growth. At the same time, successive administrations have sought to reduce reliance on semiconductor production in Asia through domestic manufacturing incentives and export controls.

Yet reshoring complex chip fabrication ecosystems remains capital-intensive and time-consuming. Despite federal initiatives, a substantial share of advanced semiconductor manufacturing capacity remains abroad, limiting the immediate domestic multiplier effect of AI infrastructure spending.

State-level regulatory initiatives aimed at governing AI development have also drawn scrutiny from some industry voices, who argue that excessive constraints could dampen investment. Others counter that regulatory clarity is necessary to sustain long-term growth and public trust.

Bubble concerns and productivity skepticism

The number of investors warning of a potential AI bubble appears to be rising. Corporate executives have acknowledged that AI is not an automatic productivity accelerant and that integrating large language models into workflows requires retraining, process redesign, and ongoing oversight.

A recurring economic concern centers on labor substitution. Some observers argue that replacing human workers with software-based systems could have second-order effects if displaced employees reduce consumption and tax contributions. While such outcomes remain speculative, they underscore the importance of distinguishing between firm-level efficiency gains and economy-wide income distribution.

Analyst Joseph Politano has suggested that AI’s macroeconomic contribution, while meaningful, has been overstated. He estimated that chatbots and large language models accounted for roughly 0.2 percentage points of last year’s 2.2% GDP growth — a fraction of headline expansion but not insignificant. However, because much of the supporting infrastructure is imported, isolating AI’s net contribution within national accounts remains challenging.

Joe Brusuelas, a tax advisor and economist, said AI’s economic effects may require future revisions as data improves. He described the current debate as an attempt to interpret incomplete signals, with analysts “trying to peer through the fog to understand what is driving growth.”

Short-term lag, long-term potential

Historically, general-purpose technologies — from electricity to the internet — have exhibited productivity lags. Significant investment often precedes measurable gains, as complementary innovations and organizational changes take time to diffuse across industries.

AI may follow a similar trajectory. Early spending is heavily concentrated in infrastructure and experimentation, while widespread productivity effects depend on integration into healthcare, finance, manufacturing, logistics, and professional services. If AI tools remain primarily assistive rather than transformative, macroeconomic gains could stay modest in the near term.

The tension between soaring equity valuations and muted GDP impact reflects this timing mismatch. Markets price expected future cash flows, while GDP measures realized output within a specific period.

What the data is suggesting is that AI’s direct contribution to official U.S. growth statistics in 2025 was limited. Thus, 2026 is expected to redefine the trajectory – erasing the bubble concern.