Trending Thursday #36

In partnership with

Hi !

This week in Silicon Valley, startups & tech:

• OpenAI breaks up with Nvidia: GPT-5.3-Codex-Spark runs on Cerebras chips, generating code 15x faster, marking a huge hit to Nvidia's inference monopoly

• AI's $200M political arms race: Anthropic drops $20M on a pro-regulation super PAC to counter OpenAI's $125M PAC ahead of the 2026 midterms

• China launches an AI blitz: MiniMax M2.5 delivers frontier intel at $0.30/1M tokens, ByteDance's Seedance 2.0 goes viral, and more

• Waymo deploys next-gen robotaxis: 6th-gen Driver cuts sensors 42%, costs under $20K per unit, and targets 1M+ weekly rides by year-end

• xAI keeps losing co-founders: Jimmy Ba and Tony Wu depart as Musk announces a reorg, while X quietly hits $1B ARR in subscriptions

As we continue to build at Context Ventures, we've launched Startup Intros with a simple mission: to help early-stage founders find the right investors, faster and smarter.

🤖 OpenAI Breaks Up with Nvidia: Codex Spark Runs on Cerebras

OpenAI just launched GPT-5.3-Codex-Spark, a lighter, faster version of its agentic coding tool. The headline number: 15x faster code generation. But the real story isn't speed. It's the chip underneath.

🔌 The Cerebras Bet

Spark is OpenAI's first model to run on Cerebras hardware, not Nvidia. It's powered by the Wafer Scale Engine 3, a megachip with 4T transistors. OpenAI calls this the "first milestone" in a $10B+ multi-year partnership announced last month.

Codex mirrors how devs work with two optimized modes:

• Spark (Rapid Prototyping): Quick iteration, low latency, and real-time collaboration run on Cerebras.

• Full Codex (Deeper Reasoning): Multi-file refactors, complex system design, and code review across entire repos run on Nvidia.

Cerebras just raised $1B at a $23B valuation and is pursuing an IPO. This isn't a science project. It's a real alternative to Nvidia's inference stack.

📉 What This Means for Nvidia

Training diversified years ago (Google TPUs, Amazon Trainium), but inference remained Nvidia's fortress. Spark cracks that wall. If Cerebras delivers lower latency at competitive cost for coding, the highest-volume AI workload, other labs will follow. Nvidia isn't threatened today (data center revenue hit $35B last quarter), but the monopoly narrative just got weaker.

Translation: OpenAI is hedging its compute future, and the first real alternative to Nvidia's inference dominance just shipped to production.

🗳️ AI's $200M Political Arms Race

Anthropic donated $20M to Public First Action, a super PAC pushing for AI guardrails and transparency. The target: counter OpenAI-backed super PACs ahead of the 2026 midterms.

💰 AI Regulation Becomes a 2026 Election Issue

• Public First Action (AI safety): Plans to raise $50M to $75M to back 30 to 50 candidates this cycle; led by former lawmakers Brad Carson and Chris Stewart.

• Leading the Future (pro-AI): Has already raised $125M from a16z, OpenAI co-founder Greg Brockman, Joe Lonsdale, Ron Conway, and Perplexity.

Total AI PAC Spending: over $200M committed for the midterms. For comparison, the entire crypto industry spent $135M on the 2024 elections. AI is already outspending crypto on political influence, and the cycle has just begun.

🎯 The Strategy: Bipartisan, Not Billionaire

Public First is backing candidates on both sides: Marsha Blackburn (kids' online safety) and Pete Ricketts (chip export limits). Their bet: 80% of Americans want AI safety rules, according to Gallup, is a stronger hand than billionaire donors. David Sacks has already criticized Anthropic for "regulatory capture."

Why This Matters: AI regulation shapes what you can build and how compliance costs look. The uncomfortable truth: the biggest companies actually benefit from regulation because they can afford the overhead. Smaller startups bear those costs disproportionately. The midterms will determine whether the US gets federal oversight or continues with a state-by-state patchwork.

Translation: The two biggest AI labs are now spending nine figures to shape the rules of their own industry, and the 2026 midterms will decide who wins.

Ship the message as fast as you think

Founders spend too much time drafting the same kinds of messages. Wispr Flow turns spoken thinking into final-draft writing so you can record investor updates, product briefs, and run-of-the-mill status notes by voice. Use saved snippets for recurring intros, insert calendar links by voice, and keep comms consistent across the team. It preserves your tone, fixes punctuation, and formats lists so you send confident messages fast. Works on Mac, Windows, and iPhone. Try Wispr Flow for founders.

Start flowing free

🇨🇳 China's AI Blitz: Four Moves in One Week

While US labs fight over politics and chips, China shipped four major moves in a single week. The "catching up" narrative seems to be dying.

💸 MiniMax M2.5: The $0.30 Revolution

MiniMax launched M2.5 at $0.30 per million input tokens and $1.20 per million output tokens. The performance is shown by the 80.2% on SWE-Bench Verified, 37% faster than M2.1, matching Claude Opus 4.6 speed. It writes specs before code, decomposes features, and plans architecture unprompted.

🛒 Alibaba's 120M Orders

Alibaba turned Qwen into an e-commerce engine. It took 120M+ orders in six days during the Lunar New Year with a $3.60 voucher. That's AI generating real commerce at scale, not a research demo. Western AI companies debate monetization while Alibaba prints transactions.

🎬 ByteDance Goes Viral

ByteDance's Seedance 2.0 video generation went viral in China, with state media calling it "bigger than Sora." The video generation race just got a serious new contender.

📈 Z.ai's Demand Problem

Z.ai is raising prices on GLM coding by 30%+ due to surging demand. When a Chinese AI company hikes prices instead of cutting them, demand outstrips supply even at rock-bottom prices.

Translation: China isn't just building competitive models; it's commercializing them faster, pricing them cheaper, and scaling them bigger than anyone expected.

🚗 Autonomy Gets Real: Waymo + Applied Intuition

🔧 Cheaper, Better, Weather-Ready

Waymo dramatically overhauled its 6th-Gen Driver on next-gen Ojai robotaxis:

• Reduces Sensors: Cameras from 29 to 13, lidars from 5 to 4, and radar to 6, to have 42% fewer sensors with greater resolution and range.

• Reduces Costs: Under $20K per unit, making it 50%+ cheaper than 5th-gen.

• Climate-Adaptable: Includes hydrophobic coatings and modular components that are swappable for different climates.

📈 1M Waymo Rides by End of 2026

Waymo currently delivers 400K+ paid rides weekly across six cities and is targeting expansion to 1M+ by the end of the year.

• Hyundai Supply: Providing 50,000 IONIQ 5 units, the largest single order in autonomous driving history.

• New Cities: 20+ additional cities planned, including Tokyo and London.

One Wrinkle: GOP lawmakers flagged concerns about Chinese-made Geely vehicles in the fleet. Waymo says it provides zero access to autonomous tech or rider data.

Translation: Autonomous driving crossed from experiment to real business at real scale, and the unit economics finally work.

🚀 Product Launches Worth Knowing

Google Gemini 3 Deep Think: Updated to better solve science, research, and engineering challenges; expanding access via the Gemini API to select researchers.

YouTube for Apple Vision Pro: Finally launched with standard, 180-degree, and 360-degree video plus Shorts. M5 chip devices play 8K.

Samsung HBM4: First commercial shipments sent, racing SK Hynix and Micron for Nvidia supply; memory bandwidth is becoming the bottleneck in AI training and inference.

Sony WF-1000XM6 Launched at $330 with AI-powered beamforming mics, improved ANC, and hands-free calling. Sony's flagship earbuds continue to set the bar.

Nvidia GeForce Now on Fire TV Native app on select Fire TV sticks, up to 1080p/60fps cloud gaming. Nvidia expands cloud gaming beyond PCs and mobile.

Amazon Kiro Internal messages reveal Amazon steering ~1,500 employees to its in-house AI coding assistant, drawing criticism from devs who prefer external tools.

Z.ai GLM Coding Hiking prices 30%+ for new subscribers to accommodate surging demand.

👔 Personnel Quick Hits: Who's Moving This Week?

xAI co-founders depart: Jimmy Ba and Tony Wu exit this week, following Igor Babuschkin, Kyle Kosic, Christian Szegedy, and Greg Yang. Six of the original co-founders are now gone.

Musk announces xAI reorg: Restructuring "required parting ways with some people" to "improve speed of execution." Meanwhile, X quietly hit $1B ARR from $3-$40/month subscriptions.

DOJ antitrust chief Gail Slater forced out: Left effective immediately after months of tensions with senior cabinet officials. Viewed as a win for tech giants over aggressive antitrust postures.

DOJ charges former L3Harris boss: Peter Williams pleaded guilty to stealing and selling eight hacking tools from Trenchant to a Russian broker with government ties.

Ben Thompson on agentic commerce: In a Stratechery Q&A, Thompson argues AI agents could reshape how consumers buy products, threatening the current ad model.

OpenAI's chief economist on adoption: Ronnie Chatterji says AI's biggest 2026 problem isn't models; it's getting companies to actually use them.

🌟 Editor's Note

At Startup Intros, our mission is to bring the latest founder-investor news straight to your inbox, keeping you ahead in the fast-paced world of Silicon Valley.

💭 Parting Thoughts

This was one of those weeks where you could feel the tectonic plates shifting.

OpenAI shipped production inference on non-Nvidia chips. Anthropic and OpenAI are spending $200M+ to fight each other at the ballot box. China made frontier intel cost $0.30 per million tokens. And Waymo proved autonomous driving scales with cheaper, better hardware.

The common thread is that labs are competing at every layer: chips, models, policy, and deployment. For founders, more competition means more options, lower costs, and more leverage. So, the best time to build is always when the giants are distracted, fighting each other.

Forward to a friend and hit reply to let me know what you're seeing in your world.

Till next time!

P.S. Raised this week, and we missed you? Want to be featured? Have tips or funding questions? Reply or DM us as we’re here to help.

Startup Intros Trending Thursday: Your trusted source for founder-investor insights, delivered with clarity and focus.