- 80/20 AI
- Posts
- ChatGPT Business Premium: Built for Ambitious Teams
ChatGPT Business Premium: Built for Ambitious Teams
Map your AI vendor relationships before NVIDIA owns the whole stack
Advertise here | 6-min Read
Turn your marketing knowledge into measurable business results.
Running more campaigns won't move the needle. Building the right strategy will.
HubSpot Academy's Digital Marketing Certification gives you the foundation to make that happen. In just over 3 hours, you'll cover consumer behavior, ethical marketing, SEO, content, email, social media, paid ads, and AI — the full stack, in one course.
Walk away with:
A complete digital marketing strategy framework
Practical skills you can apply to campaigns immediately
A globally recognized certification to add to your LinkedIn
The confidence to prove your strategy is working
Free. No credit card. Shareable on LinkedIn the day you finish.
Join 200,000+ professionals who have advanced their career with HubSpot Academy — get certified today.
ChatGPT Business Premium: Built for Ambitious Teams
AI CHEAT SHEET

OpenAI presented benchmark results for Jalapeño at Hot Chips 2026 on August 25. OpenAI's self-reported benchmarks show Jalapeño delivering 1.5x to 1.9x more AI work at peak throughput versus Nvidia Blackwell-generation systems, with 1.7x to 3.6x lower end-to-end latency and 2.1x to 4.1x faster ultra-low-latency interactive inference across GPT-OSS-120B, DeepSeek R1 and Kimi K2.5. Each rack packs 128 accelerators, 1.7 exaFLOPS of 4-bit compute, 27.5 TB of HBM4 and just under 2 petabytes per second of memory bandwidth.
The details:
The Jalapeño chip shows that a hyperscaler-designed chip can now match or beat NVIDIA's Blackwell-class GPUs on inference efficiency, Adrien Sanchez, technology analyst at Yole Group, told CNBC. He added that while NVIDIA still owns the vast majority of AI compute and has ecosystem lock-in via CUDA, OpenAI's new chip is a threat to NVIDIA's inference margins
SemiAnalysis founder Dylan Patel called the result more significant: "Usually first generation chips aren't competitive, but OpenAI is beating Nvidia Blackwell and even Rubin. This is huge news!" Gavin Baker at Atreides Management called Jalapeño "the first good ASIC outside of TPU and Trainium.
Key caveats: Jalapeño handles inference only; NVIDIA's training dominance remains untouched. The comparison targets NVIDIA's GB200 NVL72 and GB300 NVL72 racks — not the upcoming Vera Rubin platform. Tests excluded speculative decoding, and OpenAI did not disclose system-level power. Deployment trickles out later in 2026, with volume production in 2027.
Why it matters: Jalapeño is the clearest signal yet that the inference layer of AI is about to bifurcate. NVIDIA owns training — no custom chip touches it. But inference is where every user request runs, where costs pile up, and where custom silicon with a specific workload advantage can permanently change the economics. OpenAI running its own inference chips reduces its cost of goods on every ChatGPT query.
AI NEWS HIGHLIGHT
• Amazon shut Mechanical Turk after 21 years — the human-labelling infrastructure that trained the ML era is gone — from crowdsourced pennies to Mercor at $2B ARR to Google paying $10M for Spirit Airlines emails. Training data economics changed by orders of magnitude.
• Thomson Reuters launched Thomson — proprietary LLM on Westlaw + Reuters content, $40M investment — the enterprise proprietary-data playbook, executed. 150 years of legal and financial content baked into one model.
• Unitree lost 45% of its peak $66B market cap in three sessions — China IPO bubble concerns resurface — Q1 adjusted profit down 52.55%, 73.6% of robot revenue still from research/education, not industry. The humanoid robotics story is real. The valuation was not.
• CVE-2026-60004: critical Gitea vulnerability actively exploited — CISA added to KEV catalog — users with repo write access can execute arbitrary shell commands via the diffpatch API. Patch to version 1.27.1 immediately if you run self-hosted Gitea.
• OpenAI GPT-Live: native voice model, sub-300ms latency, emotional nuance — no text pipeline — the voice latency gap that made AI voice feel robotic is closing. Sub-300ms is the threshold where conversation feels real.
• Claude Sonnet 5: 5 days left at $2/$10 — September 1 is standard $3/$15. Jalapeño's volume production is 2027. The routing decision you make this week runs on today's chips and today's prices.
Build Your Proprietary AI Advantage for Your Enterprise
Prompt: You are a senior AI strategist. Thomson Reuters just trained a proprietary LLM on 150 years of Westlaw and Reuters content for $40M — not a frontier API wrapper, an owned model on owned data. Amazon shut down Mechanical Turk the same week. The training data moat is now the AI moat.
Our proprietary data: [describe what you hold — contracts, customer interactions, domain records, operational logs].
Answer four questions:
1. Data inventory — which of our data assets would give a fine-tuned model a specific, measurable advantage over a general frontier model on our most important tasks?
2. Build vs buy — compare three options for our data volume: fine-tune an open-weight model (Qwen 3.8 27B or Muse Glimmer), build a RAG system on our corpus, or train from an open-source base like Thomson did. Which delivers the most defensible advantage at the lowest cost?
3. Moat depth — how long and how much would it cost a competitor to replicate our data advantage? If the answer is "under 12 months and under $10M," we do not have a moat — we have a head start.
4. September 1 connection — for our most data-intensive workloads: would a fine-tuned open-weight model outperform a frontier API call at lower cost after September 1? If yes, the routing decision is not which frontier model — it is whether to use a frontier model at all.
End with one paragraph: what we would build, at what cost, and what advantage it creates. If the data advantage does not justify proprietary training, say so and give the RAG alternative instead.TOP TRENDING AI TOOLS
Research & Content
•Clipto— Local AI search for video, audio, meetings, and files — fully on-device, no cloud
•Perplexity AI— AI-powered search, $750M ARR, NVIDIA investment incoming — answer-first, sources embedded
•Claude— Thoughtful, context-heavy writing and deep document analysis
Software Development
•Cursor— #1 in-editor AI agent harness — use with stopping-point discipline
•Construct Computer— Your AI coworker gets a computer — multi-agent automation, scheduled jobs, persistent context
•Murmell— Cloud canvas where your team and AI agents collaborate on shared codebases
Agents & Automation
•AirJelly— Proactive, self-organising second brain — on-device memory, screen-aware, local execution
•Project SKY— Ambient AI companion for Windows — on-device memory, proactive task handling
•Wispr Flow— Dictate in any app — 4x faster than typing, writes in your voice, 100+ languages
Video & Productivity
•Soloop— AI video editor that auto-cuts, captions, and repurposes long-form content
•ElevenLabs— Industry standard for voice cloning, text-to-speech, and audio generation
That’s a Wrap
SPONSOR US
Get your business in front of over 90k+ AI professionals
8020AI is the world’s #1 AI Newsletter, Read by 90k+ professionals from leading companies such as Google, OpenAI, Meta, and Microsoft.
We've assisted in promoting Over 500 AI-Related Products. Will yours be the next?
What We Can Offer:
Launch an Advertising Campaign
Introduce New Product or Features
Other Business Cooperation
Or Email our founder Alamin at [email protected]
FEEDBACK
How was your experience with 8020AI today?
How was 8020AI today? |
If you have specific feedback or anything interesting you’d like to share, please let us know by replying to this email.







