How Silicon Valley's 'tokenmaxxing' is juicing AI demand — 4/9/2026
Speakers
Deirdre Bosa -- CNBC Anchor
— CNBC anchor, tech reporter
Summary
- The "tokenmaxxing" trend sees engineers and companies competing to consume AI tokens, which may inflate demand signals and infrastructure spending.
- Ramp CEO Eric Glyman presents data showing enterprise AI token spend is up 13x since January 2025, but no one knows how to budget for it.
- Glyman notes a "K-shaped" divergence: businesses in the top quartile of AI spend more than doubled revenue over 3 years, while the bottom quartile grew only ~12%.
- Glyman suggests AI labs like OpenAI (focused on maximizing usage/capacity) and Anthropic (more disciplined, setting limits) have contrasting strategies; he implies Anthropic's approach may be more prudent.
- Glyman raises the possibility that OpenAI faces an "innovator's dilemma" as its business model is predicated on extracting maximum spend.
- Investor Dan Niles argues the AI trade now requires differentiation, separating companies with sustainable cash flow from those without.
- Niles states OpenAI is expected to burn $220B in cash flow through 2029, creating risk for its partners and investors (e.g., Microsoft, Oracle).
- Niles identifies Anthropic as a current winner in enterprise AI, benefiting Amazon's AWS, and sees Google as resilient due to its own cash flow.
- Niles highlights a shift from GPU-centric (repetitive tasks) to CPU-centric (agentic, orchestrated tasks) compute, benefiting companies like Intel.
- A key uncertainty is how much current explosive token demand (e.g., +130% after OpenClaw's release) is sustainable versus inflated by performative usage.
Trade Ideas
Niles stated "you've seen names like Oracle get absolutely crushed because over half their backlog is related to OpenAI and OpenAI is expected to burn 220 billion in cash flow through 2029." OpenAI's massive expected cash burn and potential financial instability create direct, material risk for its major infrastructure partners, including Oracle. AVOID due to excessive concentration risk and reliance on a partner (OpenAI) with a precarious cash flow trajectory. OpenAI secures massive funding or accelerates profitability, stabilizing its operations and capital commitments.
Niles said Google is "massively free cash flow positive" and "can fund all their [AI] ambitions with the cash that they are generating organically," contrasting it with OpenAI-dependent companies. In an AI investment cycle where capital discipline and self-funding capability are becoming critical differentiators, Google's strong cash flow provides a major advantage and lower risk profile. LONG because it is a cash-flow-positive incumbent with less reliance on external funding, positioned to survive and potentially thrive through a period of potential overinvestment and shakeout. Google fails to execute on AI products and loses significant market share to competitors despite its financial strength.
Niles stated Amazon will benefit because it is "the biggest public cloud vendor" and is "hosting a lot of that Anthropic workload," and its physical infrastructure will benefit from AI/robotics in e-commerce fulfillment. Amazon's AWS is a primary beneficiary of the rise of Anthropic (a current enterprise AI leader), and its massive logistics network provides a unique, tangible use case for AI automation that can improve efficiency. LONG due to its dual-positioning as a key infrastructure host for a winning AI lab and as a major operator that can implement AI for productivity gains in its core business. Anthropic's momentum falters, or AWS loses significant cloud market share to competitors like Microsoft Azure.
Niles said he is "wary" of Microsoft because it "owns 27% of [OpenAI]" and OpenAI's cash burn and competitive squeeze create "huge issues." Microsoft's significant financial and strategic entanglement with OpenAI exposes it to the potential downside of OpenAI's financial stress and competitive challenges against Anthropic and Google. WATCH due to the high uncertainty and binary risk/reward tied to the outcome of OpenAI's financial and competitive position. OpenAI successfully navigates its cash burn and strengthens its competitive position, making Microsoft's stake and partnership highly valuable.
Niles argued that the shift to "agentic" AI, which requires orchestrating many different tasks, increases demand for CPUs (e.g., 4:1 CPU-to-GPU ratio) vs. repetitive tasks suited mainly for GPUs, benefiting "companies that have been given up for dead, like an Intel." The next phase of AI adoption (AI agents) has different compute requirements, catalyzing a resurgence in demand for microprocessor (CPU) platforms, directly aiding companies like Intel. LONG as a potential turnaround beneficiary of a shifting architectural demand within the AI compute stack, supported by recent partnership announcements (e.g., Google, Elon Musk). The shift to agentic AI is slower than expected, or Intel fails to capitalize on the demand shift due to execution or competitive issues.
