NVDA / DOSSIER
updated Aug 27, 2026The firm's house view on NVDA — written, revised, and graded by a machine.
I am holding NVDA at $209.66 rather than adding or cutting. Consensus already underwrites $105.667B of revenue and $2.4084 of EPS for the 2026-11-17 print at a ~$5,073.8B cap, which matches the CUDA lock-in I already underwrote and leaves no unused upside versus AMD and AVGO, which I also run as share-gain and custom-ASIC offsets. The tell is that 2026-11-17 print, with a leading read from MSFT, GOOGL, META on 2026-10-27 and AMZN on 2026-10-28.
Finquill's model has held a neutral stance on NVDA since July 22, 2026, and has changed its mind 0 times since coverage began on July 22, 2026.
The thesis
I think NVDA at $209.66 is a hold rather than a buy or a sell because the 2026-11-17 quarter already embeds $105.667B of revenue and $2.4084 of EPS, which is the CUDA-and-hyperscaler story I already underwrote — 80%+ accelerator share and a 15-year software lock-in — so I am not being paid for that narrative from here, only for a beat. The mechanism that keeps me from selling is that Microsoft, Alphabet, Amazon, and Meta still have to buy these accelerators, which should put a floor under that $105.667B print; the mechanism that keeps me from buying is that my AMD thesis is already share in the AI accelerator market against Nvidia and my AVGO thesis is already the custom-ASIC alternative, so adding NVDA would fight two live books and triple-count the same AI-capex dollar I also own via those four hyperscalers. The asymmetry is narrow: a beat of $105.667B could re-rate a ~$5.07T company, a miss would de-rate it faster than AMD can hedge me, and an in-line print lets time-decay eat the multiple with no expansion. I expect this to resolve on the 2026-11-17 print, with a leading read from MSFT, GOOGL, and META on 2026-10-27 and AMZN on 2026-10-28. The hold flips bullish if revenue clears $105.667B and those hyperscalers keep capex expansionary; it flips bearish if revenue or EPS misses $105.667B / $2.4084, or if AMD's 2026-11-02 call shows accelerator share coming out of NVIDIA.
What would change the machine's mind
No kill criteria pre-registered yet — an honest gap, not a hidden one. They freeze here when the desk registers them.
Changes of mind — append-only, timestamped
THIS IS WHAT THE ENGINE DOES WITH OUR BOOK.
The terminal runs the same machinery on yours — your theses captured and stress-tested, your record graded, your book watched overnight.