{"slug": "while-share-of-inference-has-held-steady-over-the-last-2-years-pre-training-has", "title": "While Share Of Inference Has Held Steady Over The Last 2 Years, Pre-training Compute Has Moved To Post-Training: Semi Analysis", "summary": "Pre-training's share of combined Anthropic and OpenAI compute fell from 67% in 1Q24 to a projected 7% by 4Q26, while post-training and reinforcement learning rose from 5% to 55%, according to a Semi Analysis chart tracking the two labs' allocation from 1Q24 through an estimated 4Q26. Inference held within a roughly 10-point band across the twelve quarters, starting at 29% and estimated at 37-38% by late 2026. Semi Analysis attributes the inversion to reasoning models such as OpenAI's o1 and Anthropic's extended-thinking Claude, which spend more compute per query, and to frontier labs having used up most freely available training data.", "body_md": "For two years, the story of frontier AI compute was simple: most of it went into pre-training, a smaller slice went into inference, and almost nothing went into what happened in between. A new chart from Semi Analysis, tracking combined Anthropic and OpenAI compute allocation from 1Q24 through an estimated 4Q26, shows that story has flipped.\n\nIn early 2024, pre-training ate up 67% of combined lab compute, post-training/RL took a sliver at 5%, and inference held 29%. By the most recent quarter in the dataset and the estimates running out to late 2026, pre-training has collapsed to single digits — 7% — while post-training and reinforcement learning have ballooned to 55%. Inference, meanwhile, has barely moved, oscillating in a band roughly between 29% and 39% across the entire two-year window, settling around the mid-to-high 30s.\n\n## The Numbers, Quarter By Quarter\n\nThe Semi Analysis chart breaks compute share into three buckets: pre-training (described as “cares about capacity”), post-training/RL (described as “mostly bandwidth-bound”), and inference (also bandwidth-bound). Plotting those three slices from 1Q24 to a 4Q26 estimate shows a near-total inversion between the first two categories.\n\nPre-training’s share fell quarter after quarter: 67% in 1Q24, 63% in 2Q24, 57% in 3Q24, and down into the 50s through 2025. By 4Q25 it had dropped to 32%, by 1Q26 to 23%, and by the 3Q26E and 4Q26E estimates it sits at just 10% and 7% respectively.\n\nPost-training/RL moved in the opposite direction almost in lockstep. It started at a negligible 5% in 1Q24, crept into double digits by 2Q25 (14%), and then accelerated sharply — 24% in 3Q25, 36% in 4Q25, 42% in 1Q26, and up to 53-55% by the 3Q26E-4Q26E estimates. In other words, more than half of all combined Anthropic and OpenAI compute is now projected to go toward post-training and reinforcement learning rather than the base pre-training runs that used to define frontier model development.\n\nInference is the one line that hasn’t moved much at all. It opened the period at 29%, peaked around 38-39% in late 2024 and early 2025, dipped into the low-to-mid 30s for much of 2025 and early 2026, and is estimated to land at 37-38% by late 2026. Across twelve quarters of data and estimates, inference’s share has stayed inside a roughly 10-point band, never collapsing the way pre-training did and never spiking the way post-training did.\n\n## Why The Shift Happened\n\nThe shift tracks closely with the industry’s broader move toward reasoning models. When OpenAI’s o1 and its successors, and Anthropic’s own extended-thinking Claude models, started spending meaningfully more compute per query to “think” before answering, the compute budget needed to teach a model how to reason well — through reinforcement learning on verifiable tasks, preference optimization, and related post-training techniques — grew far faster than the budget needed to simply pre-train a larger base model on more text. Also, the major labs have already [used up](https://officechai.com/stories/data-is-the-fossil-fuel-of-ai-it-will-get-exhausted-ilya-sutskever/) most of the data that’s freely available to train their models, so gains from pre-training would now be modest at best.\n\nThis lines up with what Semi Analysis has separately noted about frontier labs’ Tokenomics models: the compute going into post-training and inference increasingly reflects repeated reinforcement learning “hero runs” on top of a single base model, rather than one-off pre-training jobs. A lab might run a single large pre-training job and then put many multiples of that compute into post-training experiments — reward modeling, RL fine-tuning passes, and reasoning-trace generation — before a model ever reaches a customer.\n\nThat also explains why the chart labels post-training and inference as “bandwidth-bound” while pre-training is described as caring about raw capacity. Pre-training is dominated by large, dense matrix multiplications that scale efficiently with more chips. Post-training and inference, especially for reasoning models that generate long chains of thought token by token, lean much more heavily on memory bandwidth — how fast data can move in and out of a chip — rather than on how much raw compute capacity a cluster can throw at a single giant training run. That’s a big part of why Nvidia, AMD, Google and Broadcom have all been racing to improve memory bandwidth on their newest AI accelerators rather than purely chasing FLOPs.\n\n## What It Means For Anthropic And OpenAI\n\nThe practical consequence is that the two labs’ spending priorities have shifted even as total compute demand keeps climbing. With inference holding steady around a third of total compute even as post-training’s share has swelled past half, the two labs are effectively running three different compute businesses at once: a shrinking pre-training operation, a post-training operation that has become the single largest consumer of compute, and a stable but still enormous inference operation that has to serve paying customers in real time. But this will also raise the question — with inference currently requiring only 38% of available compute capacity, how will all the additional compute be used once models are at a level that any improvements would be largely indistinguishable for the average user.", "url": "https://wpnews.pro/news/while-share-of-inference-has-held-steady-over-the-last-2-years-pre-training-has", "canonical_source": "https://officechai.com/ai/while-share-of-inference-has-held-steady-over-the-last-2-years-pre-training-compute-has-moved-to-post-training-semi-analysis/", "published_at": "2026-10-18 09:51:07+00:00", "updated_at": "2026-10-18 10:46:36.508112+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-infrastructure", "ai-research"], "entities": ["Semi Analysis", "Anthropic", "OpenAI", "o1", "Claude"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/while-share-of-inference-has-held-steady-over-the-last-2-years-pre-training-has", "markdown": "https://wpnews.pro/news/while-share-of-inference-has-held-steady-over-the-last-2-years-pre-training-has.md", "text": "https://wpnews.pro/news/while-share-of-inference-has-held-steady-over-the-last-2-years-pre-training-has.txt", "jsonld": "https://wpnews.pro/news/while-share-of-inference-has-held-steady-over-the-last-2-years-pre-training-has.jsonld"}}