HP has a solution to expensive AI tokens: Buy a more expensive PC

HP argued that AI token costs can push enterprises to run models locally on “AI PCs,” a shift it says is margin accretive. Personal Systems revenue rose 18% YoY to $11.8B despite PC volumes down 16%, attributed to higher prices, premium/AI-PC mix, and attached services (now ~1/3 of Personal Systems gross profit). HP expects AI PCs to reach 60–70% of PC mix in FY2027 and >70% in 2028, while also stating it will keep increasing pricing as component costs rise; it also highlighted Precise Print, claiming up to 38% less paper and 47% lower ink use.

Analysis

This is less an “AI device” story than a pricing-and-mix story. If enterprises really start treating inference as a controllable operating expense, the endpoint vendor that can sell a higher-spec box plus security/services gets a disproportionate share of the spend, even with flat or falling unit demand. That favors HPQ’s margin structure more than pure unit-growth names, because a richer mix can offset weak PCs and push more profit into recurring attach rather than one-time hardware.

The second-order effect is a slow bleed for cloud inference economics at the enterprise margin, but I would not short AI infrastructure on this print alone. The near-term winner is the client stack: premium CPUs/NPUs, memory, storage, and device-management software all gain attachment over the next 1-3 quarters as buyers fund local compute to avoid variable usage costs. Price increases should show through immediately to a few months, so the next couple of earnings cycles matter more than any single quarter’s unit data.

The contrarian risk is that the market may over-credit the AI label while underestimating how cyclical this franchise still is. If ASPs rise but premium mix stalls or services attach fails to hold, the margin story fades quickly and the multiple should compress back toward a low-growth hardware name. Falsifiers: a sequential drop in Personal Systems margin, commentary that pricing is not sticking, or evidence that enterprises are pausing refreshes because token savings do not offset device premiums.

Key Decisions for Investors

• Long HPQ on pullbacks over the next 1-3 months; target 8-12% upside if mix/price commentary converts into guide raises, with a stop if personal systems margin fails to expand next quarter.
• Use a 2-4 month call spread in HPQ to express moderate upside while capping premium paid; this is cleaner than outright stock if the market is already partially pricing the AI-PC narrative.
• Prefer HPQ over lower-end PC exposure in the client-compute space; if you want a hedge, pair long HPQ against a basket of weaker hardware/print names that lack services attach and premium mix.
• Watch for falsification at the next quarterly update: if pricing momentum slows or AI-PC mix plateaus below the implied ramp, reduce exposure before the market re-rates this back to a cyclical.