|
Wednesday, August 26, 2026
|
|
|
|
| |
|
Today’s top story
RELEASE
OpenAI's Jalapeño chip beats Nvidia Blackwell on inference benchmarks
TechCrunch
OpenAI shared the first detailed benchmark results for Jalapeño, the custom inference accelerator it co-designed with Broadcom, at the Hot Chips conference. On SemiAnalysis' InferenceX benchmark the chip delivered more tokens per user and more throughput per kilowatt than currently available state-of-the-art inference processors, including Nvidia Blackwell systems. Hardware chief Richard Ho said a small-volume rollout is expected by the end of 2026, with broader deployment in 2027.
|
|
Why it matters
It is the first hard evidence that OpenAI's full-stack bet — co-designing models, products and silicon — can beat merchant GPUs on the economics that actually govern serving costs. Landing the same week Nvidia reports earnings, it sharpens the question of how long the largest AI buyers stay pure customers.
|
|
|
More headlines
|
|
INDUSTRY
Hugging Face reportedly weighing bids at a $13 billion valuation
Business Insider reports that Hugging Face is evaluating acquisition bids with help from banking partners at a valuation near $13 billion, roughly triple its 2023 round. No buyer has been named and no deal is final; the company earlier turned down a $500 million Nvidia investment over concerns about a single dominant shareholder.
TechCrunch
|
|
INDUSTRY
Nvidia reports second-quarter results after the close today
Analysts expect roughly $92 billion in revenue and about $85 billion from data center, which would be up more than 100% year over year. Guidance and commentary on China demand are the swing factors for a sector that treats the print as a quarterly referendum on AI spending.
Yahoo Finance
|
|
FUNDING
General Intuition raising at $6 billion, up from $2.3 billion in June
Valor Equity Partners, Point72 Ventures and Seven Seven Six are backing the spatial-reasoning foundation model startup at a $6 billion pre-money valuation, with Khosla Ventures and General Catalyst returning. Spun out of gameplay-clip platform Medal, the company is directing the capital toward robotic embodiments and more CoreWeave compute.
TechCrunch
|
|
RELEASE
Anthropic merges memory across Claude chat and Cowork
Claude's research-oriented chat and its agentic Cowork product now share one memory store, so context set in one carries into the other without a rebrief. Memories update continuously and are viewable and editable, with sensitive categories excluded by default; the feature is on across Free, Pro and Max plans.
TechCrunch
|
|
FUNDING
Stability AI raises $76 million from music labels and EA
The Stable Diffusion maker closed a $76 million Series B backed by Universal Music Group, Sony Music Group, Warner Music Group, Electronic Arts, AMD Ventures and Pacific Alliance Ventures, taking total funding to $232 million. The money goes toward its creative production suite and a larger professional services arm.
TechCrunch
|
|
This week in benchmarks
|
|
vs. Wednesday, August 19, 2026
|
|
Intelligence
Kimi K3 (max) fell from #7 to #8
|
|
Intelligence
GLM-5.3 (max) fell from #8 to #9
|
|
Intelligence
GPT-5.6 Sol (xhigh) fell from #9 to #10
|
|
Coding
Claude Fable 5 (Adaptive Reasoning, Max Effort, Opus 4.8 Fallback) climbed from #9 to #8
|
|
Coding
Claude Opus 5 (Adaptive Reasoning, High Effort) fell from #8 to #9
|
|
Intelligence
Grok 4.6 (xhigh) entered the top 10 at #7
|
|
Read past briefs and the live benchmarks at
ai-newsletter-eta-lemon.vercel.app.
Forward this to a friend who’d appreciate the brief.
|
