Skip to content
Slow Variables

Ask the data

Answers come from the same store as the site; every number is checked against the record it cites.

Indicators / series

epoch.openai.compute_spend_usd.fy

Source: Epoch AI data hub — AI companies (Epoch AI) · CC BY 4.0 · CSV

Epoch AI, 'Data on AI Companies' (CC BY 4.0), epoch.ai/data

idvaluegradeas of datepublished datetieraudited vs reportedextraction methodreview statusretrieved athttp statuscontent hashsupersedes idurlflags
e2ac42b3090bf545$416Mgrade C, evidence tier 52022-12-312023-05-045estimatedapiapproved2026-09-10T01:28:41.972289Z20090c5337eb305source page
917abc2e7cdbf458$4.0Bgrade C, evidence tier 52024-12-312024-10-095estimatedapiapproved2026-09-10T01:28:41.972289Z20090c5337eb305ce58c61ac6ce1d79source page
a30f4c282b47359a$8.0Bgrade C, evidence tier 52025-12-312026-02-205estimatedapiapproved2026-09-10T01:28:41.972289Z20090c5337eb305a45b4a116415e8desource page
Raw snippets and dispute text
  • e2ac42b3090bf545 {"Amount": "416000000", "Category": "R&D cloud compute,Inference cloud compute,Data", "Company": "OpenAI", "Confidence": "Likely", "Date": "2022-12-31", "Id": "OpenAI 2022 compute + data", "Notes": "OpenAI spend \"416M on \"computing and data\", out of ~500M in total spending.\n\n\"But it was projecting expenses of $416.45 million on computing and data, $89.31 million on staff, and $38.75 million in unspecified other operating expenses\"", "Period type": "Year", "Report date": "2023-05-04", "Source 1": "https://fortune.com/longform/chatgpt-openai-sam-altman-microsoft/", "Source Type": "Media report", "Total compute spend": "416000000.0"}
  • 917abc2e7cdbf458 {"Amount": "4000000000", "Category": "R&D cloud compute", "Company": "OpenAI", "Confidence": "Likely", "Date": "2024-12-31", "Id": "OpenAI 2024 R&D compute", "Notes": "Importantly, this may actually be ~$5-6B in incurred R&D compute expenses in 2024.\n\nThe article shows a graph with $3B in training expenses, and $1 in research compute amortization\n\nI'll take the amortization at face value here, but the actual spending on research compute in 2024 is then equal to 1B multiplied by the amortization schedule in years, which is described as multiple years. So $2B if amortized over two years, $3B if three years.\n\nThis is from a projection of 2024 spending published in October 2024, so numbers are not certain.", "Period type": "Year", "R&D compute spend": "4000000000.0", "Report date": "2024-10-09", "Source 1": "https://www.theinformation.com/articles/openai-projections-imply-losses-tripling-to-14-billion-in-2026?rc=9mzoog", "Source Type": "Media report"}
  • a30f4c282b47359a {"Amount": "8000000000", "Category": "Inference cloud compute", "Company": "OpenAI", "Confidence": "Likely", "Date": "2025-12-31", "Id": "OpenAI 2025 inference compute", "Inference compute spend": "8000000001.0", "Notes": "Per The Information: \"Last year, OpenAI spent more than $8 billion on the costs of running its AI models for its users, with roughly $4.5 billion on inference for paying users.\"\n\nThis is up ~1 billion from OpenAI's mid-2025 projection of $7B (source 2).", "Period type": "Year", "Report date": "2026-02-20", "Source 1": "https://www.theinformation.com/articles/openai-boost-revenue-forecasts-predicts-112-billion-cash-burn-2030?rc=9mzoog", "Source 2": "https://www.theinformation.com/articles/openai-spend-100-billion-backup-servers-ai-breakthroughs?rc=9mzoog", "Source Type": "Media report"}