Decision time: March 19, 2007 at 9:00 a.m. PDT. Knowledge cutoff: March 18, 2007 at 11:59:59 p.m. PDT. This learner packet contains no post-cutoff outcome evidence.
Decision
NVIDIA must choose whether to limit CUDA to graphics-adjacent support, sustain a staged external developer platform, or fund an ungated computing-platform expansion. The evidence supports staged investment, with expansion controlled by retained developers, production workloads, verified application performance, incremental hardware contribution, independent ecosystem breadth, portability, investment, and supply. judgment.nvidia.cutoff.staged-platform
This is not an invention story. The Brook for GPUs research predates CUDA and establishes general-purpose GPU programming prior art. claim.nvidia.cutoff.gpgpu-prior-art evidence.nvidia.cutoff.gpgpu-prior-art
What was observable
By the cutoff, NVIDIA had announced CUDA, paired it with GeForce 8-class hardware, and released a public beta SDK and C compiler. Contemporaneous independent reporting described both the numerical-computing architecture and the effort to attract developers. claim.nvidia.cutoff.cuda-public-toolkit claim.nvidia.cutoff.platform-intent evidence.nvidia.cutoff.independent-launch evidence.nvidia.cutoff.independent-compiler evidence.nvidia.cutoff.public-beta evidence.nvidia.cutoff.beta-toolkit
Those actions established availability and intent, not retained adoption or lock-in. The packet contains no independently reconciled active-developer, production-workload, library-usage, portability, or switching-cost series. claim.nvidia.cutoff.adoption-lockin-unknown table.nvidia.cutoff.cuda-gaps
Financial reconstruction
| Reported consolidated measure (USD millions) | FY2005 | FY2006 | FY2007 |
|---|---|---|---|
| Revenue | 2,010.033 | 2,375.687 | 3,068.771 |
| Income from operations | 95.176 | 336.664 | 453.452 |
| Net income | 88.615 | 301.176 | 448.834 |
| R&D expense | 348.220 | 357.123 | 553.467 |
| Operating cash flow | 132.2 | 446.4 | 587.1 |
| Property, equipment, and intangible purchases | (67.261) | (79.600) | (145.256) |
Every number is an as-reported consolidated fact from the cutoff filing and the structured tables. They are not CUDA revenue, cost, or return. fact.nvidia.cutoff.revenue.fy2005 fact.nvidia.cutoff.revenue.fy2006 fact.nvidia.cutoff.revenue.fy2007 fact.nvidia.cutoff.operating-income.fy2005 fact.nvidia.cutoff.operating-income.fy2006 fact.nvidia.cutoff.operating-income.fy2007 fact.nvidia.cutoff.net-income.fy2005 fact.nvidia.cutoff.net-income.fy2006 fact.nvidia.cutoff.net-income.fy2007 fact.nvidia.cutoff.rd.fy2005 fact.nvidia.cutoff.rd.fy2006 fact.nvidia.cutoff.rd.fy2007 fact.nvidia.cutoff.cfo.fy2005 fact.nvidia.cutoff.cfo.fy2006 fact.nvidia.cutoff.cfo.fy2007 fact.nvidia.cutoff.capex.fy2005 fact.nvidia.cutoff.capex.fy2006 fact.nvidia.cutoff.capex.fy2007 table.nvidia.cutoff.reported-results table.nvidia.cutoff.investment-base table.nvidia.cutoff.cash-flow claim.nvidia.cutoff.cuda-economics-unknown
FY2007 R&D included salary, acquisition, and stock-compensation effects associated with SFAS 123(R), so the increase cannot be labeled CUDA spending. claim.nvidia.cutoff.rd-comparability evidence.nvidia.cutoff.rd-accounting
Cash, cash equivalents, and marketable securities were USD 1,117.8 million at January 28, 2007, compared with USD 950.2 million at January 29, 2006. This supports a bounded learning budget, not an unlimited CUDA commitment. fact.nvidia.cutoff.liquidity.fy2006 fact.nvidia.cutoff.liquidity.fy2007 table.nvidia.cutoff.liquidity claim.nvidia.cutoff.funding-capacity
Competitive and supply boundaries
NVIDIA described competition as a combination of performance, software support, API conformity, manufacturing, and price. Its fabless model and named wafer suppliers, including TSMC, mean software effects must be separated from hardware architecture, semiconductor execution, supply, and purchase commitments. claim.nvidia.cutoff.competitive-boundary claim.nvidia.cutoff.supply-boundary evidence.nvidia.cutoff.pre-cuda-competition evidence.nvidia.cutoff.fabless-model evidence.nvidia.cutoff.wafer-suppliers evidence.nvidia.cutoff.purchase-commitments
The filing also warned that incremental R&D could be independent of revenue and that relatively fixed operating expenses might not fall quickly after a shortfall. claim.nvidia.cutoff.fixed-cost-risk evidence.nvidia.cutoff.fixed-cost-risk
Gates and abstentions
The platform should expand only after verified developer retention, production use, application performance and correctness, incremental contribution, independent complements, portability, investment, and supply gates clear. The conversion from a public toolkit to retained developer economics remains an assumption. assumption.nvidia.cutoff.developer-adoption assumption.nvidia.cutoff.platform-capture assumption.nvidia.cutoff.staged-reversibility
No cutoff-valid price, fully diluted capitalization, consensus, CUDA cash-flow series, or counterfactual is present. The packet therefore abstains from a target price, a CUDA-only return series, and shareholder-return attribution. claim.nvidia.cutoff.valuation-inputs-missing