Apple keeps AI spend flat

Apple is taking a different path through the 2026 AI spending race. A May 22 AppleInsider report said the company’s AI spend “remains flat” even as rivals and pure-play model firms keep pouring money into compute, training and infrastructure. Apple is still increasing research and development overall — CNBC reported on May 6 that R&D reached 10.3% of revenue in the March quarter — but the company has not matched the capital intensity associated with the biggest model builders. (appleinsider.com) (the-decoder.com) That matters because Apple is not trying to win the AI cycle by building the most expensive foundation stack. The company’s approach, as described across recent reporting, is to keep control of the product surface while using outside models where it thinks they improve the user experience. (appleinsider.com) ### If Apple is spending less, where is it still putting money? Apple’s March-quarter spending still rose in places that support product integration rather than hyperscale model operations. CNBC reported that Apple’s R&D expense hit 10.3% of revenue, showing the company is still funding AI work even if it is not matching the infrastructure outlays seen elsewhere in Big Tech. (appleinsider.com) That distinction is central to the story. Apple can spend on Siri, on-device features, software frameworks, chips and integration work without committing to the same training-and-serving economics that define companies whose business is selling model access itself. That reading is an inference from the spending gap described in the reporting. (appleinsider.com) ### Why does OpenAI keep showing up in this Apple story? OpenAI appears in the comparison because its economics illustrate the cost of the model-first approach. AppleInsider, citing reporting around OpenAI’s finances, said OpenAI was losing about $1.25 for every $1 of revenue; another May 22 report described adjusted operating margin of minus 122%, or roughly $1.22 lost per dollar earned. (cnbc.com) CNBC also reported in April that OpenAI had missed internal revenue and user-growth estimates while trying to fund large compute commitments. Those figures do not mean Apple has no AI costs. They do show why Apple may prefer a model in which AI is folded into devices and services, while the heaviest model-ops burden sits elsewhere. (appleinsider.com) That comparison is supported by the contrast in the reports, though Apple has not publicly framed it that way. ### What does a flat-spend posture force Apple’s teams to do differently? A restrained budget puts more pressure on orchestration. If Apple is using a mix of internal systems, on-device models and outside providers, then the hard engineering problem becomes consistency: what Siri does, how it fails, what gets handed off, and who owns the result when multiple systems are involved. That is why provider-neutral behavior contracts matter. (appleinsider.com) In practice, that means defining the user-visible rules above the model layer: what the assistant must do, what fallback looks like, what is logged, and what support teams can tell users when answers are weak or unavailable. The need for that kind of coordination is an inference from Apple’s selective integration strategy and the growing use of external AI partners in Siri-related reporting. ### Does this fit with the Siri and Gemini reports? Separate May reporting said Google’s Gemini is expected to become foundational technology for a future Siri version, while other reports have pointed to a broader Siri redesign and more chatbot-like behavior. If those reports hold, Apple’s budgeting choice becomes easier to understand: the company can preserve control over the interface and product rules while relying on outside model providers for parts of the intelligence layer. That setup would also raise the cost of cross-team confusion. Assistant teams, privacy reviewers, localization, support and release engineering would all need the same operating rules if different models can sit behind the same Apple surface. That is an inference from the architecture implied by the reports. (appleinsider.com) ### What should readers watch next? WWDC in June is the next concrete checkpoint. Apple has not laid out a new AI spending plan publicly, but upcoming Siri and Apple Intelligence announcements should show whether the company is still emphasizing product control and partner integration over brute-force model spending. (appleinsider.com) (macrumors.com)