Hyperscalers pour $400B into AI, race for next platform dominance

The race for next-platform dominance is fueling unprecedented AI investment, with hyperscalers investing hundreds of billions into the technology to control the next wave of innovation. At B2BMX 2026, analyst Benedict Evans put AI into historical perspective, comparing it to past technological shifts like mainframes, PCs, and smartphones, noting Big Four spent $400B last year on AI and budgeted $650B for 2026, with announcements of "$100 billion here... there". He argued that whoever masters this exploding capital expenditure will write the rules for enterprise software for the next decade.

Investment frenzy and shifting power

This massive capital expenditure is driven by a strategic race to build and control the next generation of computing infrastructure. By investing heavily in data centers, chips, and foundational models, hyperscalers aim to secure platform dominance in an emerging AI-driven economy, positioning themselves as indispensable utilities.

According to IMPLAN analysis, Amazon, Alphabet, Microsoft, and Meta invested $364 billion in capital expenditures during 2025 fiscal years (up from $325 billion in 2024). For 2026, Goldman Sachs reports the consensus estimate for hyperscaler capex is $527 billion, while Bloomberg reports forecasts reaching about $650 billion for new data centers and gear. These figures represent total capital expenditures, not exclusively AI spending. This aggressive spending on chips, data centers, and model training represents a significant portion of their combined revenue. While this scale allows hyperscalers to absorb costs that would crush startups, investors are questioning the return on this infrastructure. Evans warned that the science remains fluid, meaning today's front-runner could stumble if a leaner architecture emerges.

B2BMX 2026: Benedict Evans Explores the Next Major Platform Shift & Business Impacts of AI

Evans delivered a pragmatic message for B2B marketers: AI is no longer a research project but an active tool optimizing ad copy, sales outreach, and lead scoring. The next frontier is "agentic" systems - AI that can execute complex tasks across multiple applications without human intervention. As software and AI merge, users will become indifferent to whether a process is driven by code or cognition.

To navigate this new landscape, Evans recommended that businesses:

• Treat AI models like cloud services - focus on capacity, not brand names.
• Identify tasks requiring human judgment and prioritize automating everything else.
• Monitor latency, as processing speed is becoming a critical conversion factor.
• Integrate prompt engineering skills into all demand generation teams.

He predicted the market will consolidate around a few dominant global models, much like the iOS and Android duopoly in mobile operating systems. Smaller AI companies will likely survive by securing distribution through cloud providers or leveraging unique, proprietary data.

Labor market ripple effects

Addressing concerns about job displacement, Evans referenced the World Economic Forum's most recent Future of Jobs Report 2025, which projects that automation will displace 92 million jobs while creating 170 million new roles by 2030. The critical challenge for business leaders is managing this transition. Entry-level administrative roles are most at risk, while demand is surging for specialized roles like prompt engineers, AI ethics officers, and data-center specialists.

Evans urged companies to invest in reskilling all employees, not just technical teams. For instance, a marketing analyst skilled in crafting sophisticated prompts can significantly boost campaign ROI, creating value that more than offsets pressure on headcount in other areas.

What consolidation means for marketers

As the AI market consolidates, a few hyperscalers will likely control access to the most powerful models, giving them significant pricing power. Evans advised businesses to lock in multi-year token contracts now to hedge against future price hikes. He also advocated for pushing industry-wide interoperability standards to prevent vendor lock-in and maintain manageable switching costs.

The key takeaway for B2B teams is that AI is neither a silver bullet nor a passing trend. It is the next computing platform, arriving faster and with a higher price tag than any before. Success will belong to those who adapt to its volatility, not those who wait for a guaranteed outcome.

---

What exactly is the projected $400B+ that hyperscalers are investing in AI infrastructure?

The figure covers the combined capital expenditure of the four largest tech platforms - Amazon, Alphabet, Meta and Microsoft - on data-center builds, custom AI chips, cooling systems and the specialized networking gear required to train and run giant foundation models. In his B2BMX keynote Benedict Evans noted the spend is "driven by FOMO" rather than by any proven path to margin; the cash is going into concrete, steel and silicon long before revenue models are clear.

How does this wave compare with earlier platform shifts?

Evans frames AI as the fourth major cycle after mainframes → PCs → smartphones, but with a twist: the science is still changing under our feet. Each prior shift took 10-15 years to reach mass adoption; AI models are being re-architected every six months, so the investment curve is running ahead of the adoption curve in a way we have not seen before.

Will the LLM market consolidate to only a few winners?

Consensus from the capex breakdown is that only hyperscaler-backed models can afford the next round of trillion-parameter training runs. Evans expects the field to narrow to "two or three foundation-model franchises" that sell API access to everyone else, turning today's 20-plus LLM vendors into a handful of utilities - similar to how cloud computing distilled to AWS, Azure and GCP.

Where is value being captured today - and where is it leaking?

Near-term revenue is landing in semiconductors (NVIDIA, custom ASICs) and data-center REITs, not in the models themselves. Evans points out that advertising, enterprise SaaS and agentic commerce are the three surfaces where AI can add dollars faster than it burns them; everything else is still an R&D line item.

What does the spending spree mean for jobs inside B2B organizations?

AI is already reshaping marketing and sales stacks - think automatic lead scoring, synthetic SDRs, real-time content variants - but Evans stresses it is "a normal technology that will take time." Short term, expect productivity gains of 15-20 % in content-heavy roles; medium term, new job families such as prompt ops, model governance and AI revenue analytics offset classic displacement in data-entry or inside-sales tiers.