汇丰 HSBC：全球科技行业研究：巨额业务投入 OpenAI能否承担？-251017（英文版）（32页）.pdf

1、 Disclosures&Disclaimer This report must be read with the disclosures and the analyst certifications in the Disclosure appendix,and with the Disclaimer,which forms part of it.Issuer of report:HSBC Continental Europe View HSBC Global Investment Research at:https:/ The 21stedition of the EM Sentiment

2、SurveyClick to viewBulls on the march OpenAI needs an extra 32GW computing power that costs USD1.6tr to build,but less so to rent for OpenAI LLM TAM of USD500bn in 2030e;we forecast OpenAIs P&L and cash flow,and its ability to meet chipmaker expectations We recommend exposure to OpenAI through NVIDI

3、A,AMD,Broadcom,TSMC,and Microsoft,all rated Buy Chips are on me:Significant compute-for-equity deals have been announced between OpenAI and tech firms.We calculate that these deals aim to provide OpenAIs current c2GW with 32GW of additional computing power.Chip makers aim to provide support to OpenA

4、I,from legacy firms like NVIDIA to AMD,to Arm and Broadcom.Infrastructure owners are involved too,from Microsoft(a historical investor in OpenAI)to Oracle and CoreWeave.Some deals include equity injection(NVIDIA)or shares(AMD).Vendors financing the expansion of their most promising client is not nec

5、essarily bad,but the scale of interlaced engagement will require solid revenues from AI large language models(LLM).We calculate that deploying 1GW costs USD50bn,hence the total bill could reach USD1.6tr.But OpenAI is not exposed to the same extent it will rent the deployed capacity,not build.Based o

6、n our estimated schedule for computing power release,we calculate that OpenAI can cover its future computing costs of cUSD500bn by 2030.TAM of USD500bn by 2030e and OpenAI as a leader:We are positive about the monetisation prospects of AI LLMs.We present our forecasts for the total addressable marke