# NVIDIA

[https://timesof.ai/platform/nvidia](https://timesof.ai/platform/nvidia)

NVIDIA provides the dominant GPU and software infrastructure powering modern AI training and inference.

NVIDIA's hardware and software ecosystem underpins the majority of frontier AI development.

## Coverage

- [Samsung Guides to ₩107.4 Trillion Quarterly Profit as Analysts Credit AI Memory](https://timesof.ai/2026/10/samsung-q3-2026-guidance-107-trillion-won-hbm4) — Samsung guides Q3 operating profit to ~₩107.4T, a Korean record. Analysts credit HBM4 memory; the division split is estimates until Oct. 29 results.
- [RTX Spark Gets a $2,599 Price, and Windows Gets a Cage for Agents](https://timesof.ai/2026/10/microsoft-nvidia-rtx-spark-surface-laptop-ultra-mxc-ga) — Microsoft prices the first RTX Spark PC at $2,599 and makes Execution Containers for AI agents generally available. Specs are vendor claims.
- [SpaceX In Early Talks for ~$40B Debt, Led by Apollo, to Buy Nvidia Chips](https://timesof.ai/2026/10/spacex-40b-debt-nvidia-chips-apollo) — FT and Bloomberg: SpaceX in early talks to borrow ~$40B for Nvidia chips, Apollo leading. Early stage; may not close. Borrower is SpaceX.
- [Lambda Targeting Up to $4B at $14.5B Pre-Money Ahead of Planned 2027 IPO, WSJ Says](https://timesof.ai/2026/10/lambda-4b-pre-ipo-round-14-5b) — WSJ: Lambda targeting up to $4B at $14.5B pre-money, Blackstone and Coatue leading, 2027 IPO aim. Backlog $50B. Single-origin; not yet raised.
- [South Korea Plans a ₩4.7 Trillion (~$3.5B) State Equity Push for a Frontier AI Model in 2027](https://timesof.ai/2026/10/korea-frontier-ai-project-4-7t-won) — Ryu Je-myung: Korea to launch state equity frontier AI project in 2027. Korea Times: ₩4.7T earmark (₩3.9T for 10k Vera Rubin GPUs, ₩800B data). Match funding required; not yet funded.
- [Update: Firmus’s IPO Documents Reportedly Pulled From Investors’ Data Room as Bankers Weigh Shrinking or Scrapping the Float](https://timesof.ai/2026/10/firmus-ipo-half-to-existing-investors) — Update: Firmus’s prospectus was pulled from the investor data room and bankers reportedly weighed scrapping the float. Firmus has made no announcement.
- [AM Intelligence Places Binding Orders for 20,000 Nvidia Rubin GPUs — ~70 MW in India and Malaysia](https://timesof.ai/2026/10/am-intelligence-20k-rubin-gpus-india-malaysia) — AM Intelligence: binding orders for 20,000 Nvidia Rubin GPUs (~70 MW India/Malaysia, Q2 2027). ~29k GPUs committed. ~$4B is Mint’s source figure, not in the release.
- [Update: Reflection Shows Beam — 501B / 23B Active, Apache 2.0 Weights Later This Month](https://timesof.ai/2026/10/reflection-open-weight-imminent) — Reflection unveils Beam (501B/23B active MoE). Self-reported benches near GLM-5.2; Kimi K3 and GLM-5.3 still ahead. Apache 2.0 weights later this month. Not on AA yet.
- [Nvidia DGX Spark 64GB: $4,999 Local AI Box On Sale Oct. 23](https://timesof.ai/2026/10/nvidia-dgx-spark-64gb) — Nvidia DGX Spark 64GB starts at $4,999 on Oct. 23 from major OEMs; up to 100B params on device. 1.7x cluster speedup is Nvidia’s Qwen 3.8 27B test.
- [Trump and Frontier CEOs Sign a Voluntary ‘Morally Binding’ Accord — No Legal Force](https://timesof.ai/2026/09/white-house-frontier-responsibilities-accord) — Trump and frontier CEOs signed a voluntary Joint Commitment on Frontier Responsibilities: four-layer self-audits, no legal force, company-chosen auditors.
- [Salesforce Owns a CRM Model — Koa on Nemotron, Weights Inside the Trust Boundary](https://timesof.ai/2026/09/salesforce-koa-nvidia-nemotron) — Koa is Salesforce’s CRM reasoning model post-trained from Nemotron 3 Super. Weights stay inside Salesforce; 3× fewer errors is vendor-only.
- [Lehane Says Safety Talks With Anthropic and Google Are Live — No Antitrust Waiver Needed](https://timesof.ai/2026/09/openai-lehane-anthropic-google-safety-talks) — Chris Lehane says OpenAI, Anthropic, and Google have talked AI safety for weeks without needing an antitrust waiver. Huang and Zuckerberg push back.
- [TCS, Not a Hyperscaler, Commits ~$7.4B for a 1 GW Hyderabad AI Campus](https://timesof.ai/2026/09/tcs-hypervault-hyderabad-1gw) — TCS HyperVault commits ~$7.4B for a phased 1 GW liquid-cooled AI campus in Hyderabad. India infra play from an IT services giant, not a hyperscaler.
- [Nvidia Buys the Model Supermarket — Hugging Face for $12.93B](https://timesof.ai/2026/09/nvidia-acquires-hugging-face-12-9b) — Nvidia’s definitive $12.93B Hugging Face deal targets H1 2027 close. Hub promised multi-hardware; regulators and open-weight labs are the watch items.
- [Nvidia Is Lessor, Not Just Chip Vendor, on Anthropic’s Reported $35B Texas Block](https://timesof.ai/2026/09/anthropic-lambda-35b-hut8-nvidia) — Reported $35B Anthropic–Lambda cloud deal at Hut 8 Beacon Point (~350 MW). Nvidia holds the lease; Anthropic declined comment. Not Nscale WV.
- [Nvidia Finances MediaTek So Custom XPUs Still Plug Into Nvidia Racks](https://timesof.ai/2026/08/nvidia-mediatek-3-5-billion-nvlink-fusion) — Nvidia bought $3.5 billion of MediaTek convertibles and opened NVLink Fusion for custom XPUs that drop into Nvidia rack-scale factories.
- [Nvidia’s Q2: Rubin Already One-Fifth of Data Center, $108B Guide, China-Zero](https://timesof.ai/2026/08/nvidia-q2-fy2027-96-billion) — Nvidia printed $96.2 billion Q2 revenue, $108 billion Q3 guidance with no China DC compute, and Vera Rubin already ~20% of data-center sales.
- [AWS Buys Another 2 Million Nvidia GPUs — And Wires Trainium Into NVLink Fusion](https://timesof.ai/2026/08/aws-nvidia-2-million-gpus) — AWS adds two million more Nvidia GPUs in 2027–2028, plus Vera CPUs and 100,000 government GPUs — Trainium on NVLink Fusion is the interesting clause.
- [Anthropic’s Reported $45B Nscale Lease Is a State-Sized Power Block](https://timesof.ai/2026/08/anthropic-nscale-45-billion-west-virginia) — Bloomberg: Anthropic to pay Nscale about $45 billion for 460 MW of Vera Rubin compute in West Virginia from late 2027. Lab declined comment.
- [Jalapeño Posts Pareto Numbers on Open Models — OpenAI Is No Longer Only a Nvidia Tenant](https://timesof.ai/2026/08/openai-jalapeno-inference-chip-hot-chips) — OpenAI published Jalapeño InferenceX results: 1.5–1.9× more work per watt and 1.7–3.6× lower latency versus Nvidia GB200/GB300 — lab-reported, not a customer SKU.
- [Taiwan Indicts Nine — Including an Nvidia Manager — Over B300 Exports to China](https://timesof.ai/2026/08/taiwan-nvidia-b300-smuggling-indictment) — Taiwan prosecutors indicted nine people, including an Nvidia Taiwan manager, over alleged B300 AI server exports to China — indictment, not conviction.
- [xAI Takes Vera CPUs for Agents — and Plans a Vera Rubin Rack in Orbit](https://timesof.ai/2026/08/spacexai-nvidia-vera-cpu-starmind) — xAI will deploy Vera CPUs for Grok agents and fly an optimized Vera Rubin NVL72 as Starmind — Nvidia primary; Starmind planned, not launched.
- [Groq 3 LPX Is a Rack You Can Order — AA Hit 3,431 tok/s on Nvidia’s Own Gear](https://timesof.ai/2026/08/nvidia-groq-3-lpx-full-production) — Nvidia put Groq 3 LPX into full production; Artificial Analysis hit 3,431 tok/s on Gemma 4 31B at a private Nvidia-hosted endpoint. Nebius first.
- [Nvidia’s Perplexity Talks Are Equity — With a Poolside-Shaped License Option Still on the Table](https://timesof.ai/2026/08/nvidia-perplexity-30-billion-investment-talks) — The Information: Nvidia discussing a Perplexity investment at >$30B valuation — talks, not signed; license-plus-hire was also considered.