Artificial Intelligence
2026: The AI Inflection Year — Why Wedbush, Nvidia and Alphabet Are All Saying the Same Thing
Wedbush calls 2026 the AI inflection year. Nvidia’s revenue grew 73% YoY. Alphabet’s cloud surged 48%. Here’s the investor roadmap for the most consequential year in AI history.
Wall Street’s most vocal AI bull, Wedbush Securities analyst Dan Ives, has been saying it for months: 2026 is not just another year in the artificial intelligence investment narrative. It is the inflection point — the year AI moves from promise to profit at scale, from infrastructure buildout to revenue generation, from narrative trade to fundamental story. The data from Q4 earnings season is starting to prove him right.
The Numbers That Define the Moment
Nvidia reported revenue of $68.1 billion in the fourth quarter of its fiscal year 2026 (ended January 25) — a 73% year-over-year increase. To put that in context: Nvidia is now generating revenue at an annualized rate exceeding $270 billion per year, growing at roughly three times the pace of most large-cap technology companies at their peak growth phase.
The Blackwell platform, which powered much of this growth, is being succeeded by the next-generation Rubin architecture, expected to roll out in the second half of 2026. Each successive generation of Nvidia’s compute architecture has expanded the addressable market rather than simply cannibalizing the prior generation — a pattern that suggests the ceiling for AI infrastructure spending is far higher than consensus estimates currently reflect.
Alphabet delivered 48% year-over-year growth in cloud computing revenue in Q4 — a business scaling dramatically faster than its consumer search operation, which itself remains the world’s most profitable digital advertising business. The combination of AI-native product development across search, cloud, enterprise software, and hardware makes Alphabet one of the most diversified plays on the AI secular trend.
Hyperscaler capital expenditure, first quarter
The Capex Says This Is Not an Inflection Year
An inflection from buildout to profit has a signature in the data: capital expenditure growth decelerates while revenue growth holds up. That is what “moving from infrastructure to monetisation” means arithmetically. It is worth checking whether it happened.
Our hyperscaler capex tracker reads combined purchases of property and equipment straight from Microsoft, Alphabet, Amazon and Meta’s SEC filings. Because the four run different fiscal years, we compare first calendar quarters only — the one period all four report on the same three-month basis.
| Q1 | Combined capex | YoY growth |
|---|---|---|
| 2023 | $33.9bn | — |
| 2024 | $44.3bn | +30.7% |
| 2025 | $71.9bn | +62.3% |
| 2026 | $129.8bn | +80.5% |
Capex growth did not decelerate. It went 30.7% → 62.3% → 80.5%, accelerating in each of the last three years. Q1 2026 alone — $129.8 billion — exceeded the combined first-quarter capital expenditure of these four companies across the entire 2018–2022 period ($113.4 billion).
That is not an inflection point. It is the steepest part of a capital cycle.
The distinction matters for positioning. Nvidia’s 73% revenue growth is not evidence that AI has begun paying for itself; it is a measurement of how fast its customers are spending. Those two facts are the same fact, viewed from opposite sides of an invoice. Combined hyperscaler capex grew at 80.5% over roughly the same window — faster than the revenue it produced at the largest supplier. The buildout is outrunning its own supply chain, which is the classic profile of a boom’s middle rather than its resolution.
None of this makes the spending irrational. It makes the “inflection” framing premature. What 2026 actually shows is maximum commitment, not proven return.
The test that would prove the thesis
A genuine inflection would look like this, and each part is observable:
- Capex growth decelerates — say from 80% to 20–30% — while hyperscaler cloud revenue growth stays in the 40s. That is the buildout beginning to be absorbed rather than extended.
- Operating margins hold as depreciation catches up. Everything in that table lands on the balance sheet and then runs through the income statement over the asset’s life. While capex compounds at 80%, depreciation lags badly and margins flatter the business. At a stable run rate, annual depreciation converges on annual capex. The margin cushion is a function of the growth rate, not the franchise.
- Application-layer revenue scales without a matching increase in the capex line. That is the actual definition of AI paying for itself, and it is the piece with the least public evidence today.
Until at least the first two show up in filings, the correct description of 2026 is an infrastructure year of unprecedented scale. That is a good environment to own. It is a different environment from the one the “inflection” label describes, and it fails in a different way.
Training vs. Inference: The Market Is Splitting
One of the most consequential developments in the AI investment landscape in early 2026 is the emerging bifurcation between the training market and the inference market — and understanding this distinction is essential for investors.
The AI training market — building and improving large language models — has exploded over the past three years. It has been cyclical, capital-intensive, and dominated by a small number of hyperscale customers. But the inference market — the actual deployment of AI models to serve end users in real time — is where the next phase of growth is concentrated, with very different economics: continuous workloads, low-latency demands, and cost efficiency over raw compute power.
Nvidia’s $1.5 billion investment in Groq, a specialized inference chip company, signals clearly where the smart money sees the next hardware battleground. Broadcom, which designs custom AI accelerators for Google and Meta’s inference workloads, is another name drawing significant institutional attention. The investment implication: portfolios overweight pure training-cycle plays relative to inference infrastructure may be positioned for the last cycle rather than the next one.
The $3–8 Trillion AI Infrastructure Investment Wave
Morgan Stanley Research estimates nearly $3 trillion in AI-related infrastructure investment will flow through the global economy by 2028 — with more than 80% of that spending still ahead. BlackRock’s Investment Institute goes further, projecting an additional $5–8 trillion in AI-related capex through 2030. These numbers encompass not just chips and data centers, but the full technology stack: networking, cooling systems, power generation, software platforms, data management, security, and the human capital required to deploy all of it.
Beyond the Obvious: The Sectors to Watch
Power and utilities. AI data centers consume extraordinary amounts of electricity. The demand for reliable, low-carbon power is creating a renaissance for nuclear energy companies, utility-scale solar developers, and grid infrastructure providers. The AI boom may prove to be the single biggest catalyst for energy infrastructure investment in a generation.
Enterprise software. As AI capabilities become embedded in every major software platform — from CRM to supply chain management to financial analytics — the companies that successfully integrate AI into their core product offerings will see significant competitive moat expansion.
Cybersecurity. Every AI system is a potential attack surface. The proliferation of AI across critical infrastructure, financial services, and government systems is driving sustained demand for AI-native security solutions.
The Skeptics and the Bubble Question
Any honest investment analysis must acknowledge the skeptical case. AI valuations, particularly for pure-play infrastructure names, are not cheap by any traditional metric. The question of whether AI spending is generating sufficient economic returns remains genuinely open. However, the key distinction for 2026 is that AI is already generating real revenue at scale. This is not 1999. The companies building and deploying AI are generating cash flows that justify sophisticated valuation frameworks, even if those frameworks require longer time horizons than typical equity analysis.
FAQ
Is 2026 really the AI inflection year?
Not by the test that phrase implies. An inflection from buildout to monetisation would show capital expenditure growth slowing while revenue growth held. Instead, combined Q1 capex at Microsoft, Alphabet, Amazon and Meta accelerated from +30.7% to +62.3% to +80.5% over the last three years, reaching $129.8 billion in a single quarter. 2026 is the largest infrastructure year on record for these companies. Whether it pays is a later question.
How much are the big four spending on data centres?
$129.8 billion in the first quarter of 2026, against $71.9 billion in Q1 2025, taken from their own SEC filings. This is total property and equipment spend — no filing discloses an AI-only figure, and we do not estimate one. Even so, four times a single quarter puts these four alone above $500 billion a year before counting Oracle, the neoclouds or the sovereign programmes.
Is this an AI bubble?
The spending is contracted, disclosed and verifiable, which distinguishes it from a purely narrative mania. The open question is return on that capital, and it is genuinely open: capex compounding at 80% a year has to be serviced by recurring revenue that is currently a projection rather than a disclosure. The useful discipline is not to settle the bubble question in advance but to watch for the deceleration and the depreciation catch-up, both of which appear in filings.
What is the difference between the training and inference markets?
Training builds and improves models — cyclical, capital-intensive, concentrated in a handful of hyperscale buyers. Inference runs those models for end users — continuous workloads, latency-sensitive, and judged on cost per query rather than raw compute. They favour different silicon and different suppliers, which is why Nvidia’s $1.5 billion investment in Groq and Broadcom’s custom accelerator business matter more than their revenue currently suggests.
The Bottom Line for Investors
AI is generating real revenue at scale, and the infrastructure wave is still largely in front of us. But the “inflection year” label describes a transition the data has not yet made: capex is accelerating, not decelerating, and the application layer is at the beginning of monetisation rather than the middle.
For investors, the practical consequence is that this remains a capital-cycle trade, not yet a cash-flow trade. Structure it accordingly — balance hyperscaler names against less obvious infrastructure plays, position for inference alongside training, size for the drawdowns that accompany any secular theme, and watch the capex tracker for the quarter the growth rate breaks. That number will call the turn earlier than any research note.
Stay ahead of the markets. — AI Capital Wire Team