AKAM · Q4 2025 Earnings

Narrative arc: CIS inflection thesis (Q2) → Akamai Inference Cloud platform launch (Q3) → AI inference as primary growth engine with $450M committed capital (Q4).

Three quarters ago Akamai talked about cloud compute as a 15% growth opportunity that might miss its target. Two quarters ago it launched Akamai Inference Cloud at NVIDIA GTC as a forward-positioning bet. This quarter, in Q&A with Sanjit Singh (Morgan Stanley), Tom Leighton said "the inference cloud offering we announced in the fall where we deployed the GPUs into 20 cities, that's already sold out, even though it's not generally available yet... we intend to invest approximately $250 million of CapEx this year to augment our AI inference cloud." The shift from theoretical opportunity to sold-out product requiring emergency capacity expansion in two quarters is the most aggressive narrative escalation Akamai has run in years.

Capital intensity went from "holding steady" to "supply-shock plus inference build-out." Last quarter capex was guided at 16% of revenue for Q4 and 20% for the FY (FY25 actual came in at 19%). This quarter management disclosed: "we are seeing a dramatic increase in the price of memory chips, which is driving up the cost of servers. This supply constraint has necessitated an upward adjustment to our CapEx forecast of approximately $200 million for 2026." They were unusually candid that "excluding the impact of the increased hardware pricing, 2026 CapEx would have trended within the 18% to 22% range" — separating transitory inflation from structural investment. Investors should still discount the bridge: even at the normalized 18–22% range, capex is at or above FY25 actuals of 19%, and the chip-inflation framing assumes pricing reverts.

The reporting framework change makes the strategic pivot official. "For the compute business, we will begin reporting it as a standalone revenue category effective in the first quarter of 2026." This is the corporate-disclosure equivalent of a thesis change. Akamai no longer wants to be read as security-plus-delivery-plus-emerging-compute; it wants to be read as AI infrastructure with two legacy businesses funding the build-out. Worth flagging because the same management team withdrew the explicit FY security revenue guide last quarter and pivoted to ARR — disclosure changes have been consistently directional toward the AI/CIS narrative.

Security framing softened materially on the blended number. Last quarter the explicit ~10% constant-currency FY security revenue guide was withdrawn. This quarter, FY26 security is guided to "high single digits in constant currency" — a step down from the implicit 10–11% trajectory of FY25. Management is leaning harder on the high-growth security ARR bucket (Guardicore + API at +36%, with API alone >100% YoY) and accepting that the blended number drifts to 7–9%. The bifurcation isn't new; the willingness to guide the blended number lower is.

Delivery narrative was validated by the print, not the rhetoric. -2% Q4 vs. -4% Q3 is the first sequential improvement after two quarters of widening decline. But FY26 guide for "delivery and other cloud apps" is "decline significantly in the mid-single digits year over year" — which actually implies a re-acceleration of decline from FY25's actual delivery print. The framing is less bullish than Q3's AI-driven-video-enrichment optimism suggested. Trough may be in for Q4, but FY26 trajectory remains down.

Confidence reads as 5/5 on the AI thesis and 3/5 on the FY26 P&L. Management was forthcoming in Q&A — including the disclosure that the $200M anchor is an existing CDN/security customer, that a rental-by-the-hour GPU service launches this quarter with a list price of $250, and that the early inference pipeline skews toward customers seeking reserved-capacity multi-year commitments rather than on-demand consumption.