Community Discussion · Policy

WestTide · AI · 2026-08-13 · Issue 35

West TideAug 132026/08/12 280 views

WestTide · AI · 2026-08-13 · Issue #35

Today's Highlights: Nvidia teams up with Wall Street to raise $500 billion for AI infrastructure, turning compute into a tradable asset; Google's Gemini app surpasses 1 billion monthly active users, joining ChatGPT in the era of "mass-market apps"; Anthropic's unreleased new model pushes the mathematical record on the Riemann Hypothesis from 41.6% to 67.2%; Chinese models sweep the top four spots on OpenRouter's weekly leaderboard.

Editor's Note: The signal this week is the "financialization of compute." When chip companies start borrowing from Wall Street to buy machines, the next round of the AI race shifts from a technological battle to a capital war.

I. AI Large Models (LLM / Foundation Model)

1. Nvidia Teams Up with Wall Street: Raising $500 Billion for AI Infrastructure

Summary: Nvidia is collaborating with Wall Street to raise $500 billion in third-party capital for AI data centers via Special Purpose Vehicles (SPVs). Jensen Huang aims to turn AI compute into financeable infrastructure assets like highways and power grids, where investors earn returns based on compute rental revenue. The news sparked both excitement and concern—Morgan Stanley warned of AI compute bottleneck risks, while some commentators called it "a massive gamble."
Source: Bloomberg · 2026-08-13
Editor's Take: AI is moving from "buying chips" to "renting/selling compute." For ordinary people, this means the huge sums spent by giants will eventually be passed down through cloud service prices, making short-term reductions in AI usage costs unlikely.

2. Gemini Becomes Google's Fastest-Growing Product Ever: MAU Surpasses 1 Billion

Summary: Google CEO Sundar Pichai announced that the Gemini app has surpassed 1 billion monthly active users, becoming the fastest-growing product in Google's history. Gemini achieved in under two years what took Google Search over a decade, joining ChatGPT as the only two AI assistants with 1 billion users.
Source: TechCrunch · 2026-08-12
Editor's Take: 1 billion users is a milestone, but also a countdown to monetization. The free era won't last long; both Google and OpenAI need to convert traffic into revenue.

3. OpenAI Launches ChatGPT Linux Desktop Version

Summary: OpenAI released a preview version of the ChatGPT desktop app for Linux, completing the final piece of its developer ecosystem. Linux is the primary system for programmers and AI engineers; previously, the ChatGPT desktop app was only available on macOS and Windows.
Source: Hacker News · 2026-08-13
Editor's Take: Developers are the most hardcore user base for AI. This move by OpenAI aims to make "using ChatGPT to write code" a daily routine for programmers, competing for talent against Claude Code and Cursor.

4. Anthropic's Unreleased New Model Breaks Math Record: Riemann Hypothesis Progressed to 67.2%

Summary: Anthropic challenged the Riemann Hypothesis with an unreleased new model. Although it didn't solve it, it pushed the lower bound of the "proportion of zeros on the critical line" from 41.6% (a figure reached by the math community after 37 years) straight to 67.2%. This is another preview of AI's capabilities in mathematical reasoning.
Source: Hacker News · 2026-08-12
Editor's Take: There is a chasm between "close to proof" and "proof," but this number indicates that AI's mathematical reasoning is rapidly approaching top human levels. The future form of mathematical tools will be rewritten.

5. Anthropic Publishes Explanation on How Claude Watermarks AI Content

Summary: Anthropic released a technical note explaining how Claude attaches invisible watermarks to generated text and image content to comply with the EU AI Act's content labeling obligations. These watermarks are invisible to regular readers but can be traced using technical methods.
Source: Daring Fireball · 2026-08-13
Editor's Take: Tracing AI-generated content is becoming an industry standard. In the future, "Is this written by AI?" will be a technical question rather than a guess.

6. OpenRouter Weekly Leaderboard: Chinese Models Exceed US in Weekly Calls for 15 Consecutive Weeks, Sweeping Top Four

Summary: Latest weekly data from OpenRouter shows that total global AI large model calls reached 69 trillion tokens from August 3 to 9, a 21.48% increase week-over-week. The top four spots were swept by Chinese models: DeepSeek-V4-Flash official version debuted at #1 (8.83 trillion, a 570% surge), Tencent Hy3 at #2 (8.05 trillion), DeepSeek-V4-Flash preview at #3, and Xiaomi MiMo-V2.5 at #4. Chinese models' weekly call volume hit 34.25 trillion, exceeding the US's 9.17 trillion for the 15th consecutive week.
Source: Guancha.cn · 2026-08-10
Editor's Take: Developers vote with their feet. Open source + cheap + easy to use has allowed Chinese models to reverse the offense-defense dynamic in the global API market, which is more convincing than any launch event.

II. AI Software (AI Application / SaaS)

1. AI Coding Firm Cognition Negotiating New Funding: Valuation Soars to $40 Billion

Summary: Cognition, maker of the AI coding assistant Devin, is reportedly negotiating a new funding round that could jump its valuation from billions earlier this year to $40 billion. Devin is a star product in the "autonomous coding" track.
Source: TechCrunch · 2026-08-12
Editor's Take: AI coding is the hottest track for capital—code is the "infrastructure" of the digital world. Whoever lets AI write clear code holds the key to the next generation of development entry points.

2. Lovable Confirms $13.3 Billion Valuation, Raises Another $400 Million

Summary: European "hands-off coding" star Lovable confirmed a new $400 million funding round at a $13.3 billion valuation. Lovable focuses on generating applications directly from natural language descriptions and is seen as a representative of the "writing software by talking" wave.
Source: TechCrunch · 2026-08-13
Editor's Take: People who don't know how to code are becoming "developers." The hotter these tools get, the more precarious the positions of traditional software outsourcing and junior programmers become.

3. xAI Co-founder's Startup River AI Lands $1.1 Billion Funding Just Two Months After Founding

Summary: River AI, founded by xAI co-founder Igor Babuschkin, secured $1.1 billion in seed/Series A funding led by General Catalyst. The company is only two months old and targets general-purpose AI agent infrastructure.
Source: TechCrunch · 2026-08-12
Editor's Take: Two months, $1.1 billion—the names of top AI talents are valuations themselves. Capital's hunger for "the next OpenAI" is evident.

4. Accel Raises $3.5 Billion, Betting Big on Global AI Startups

Summary: Veteran VC Accel announced a new $3.5 billion fund focused on AI startups in emerging markets globally. Previously, Accel completed a $550 million India fund within weeks.
Source: Hacker News · 2026-08-13
Editor's Take: Money is flocking to AI, but valuations are being quickly overdrawn. Investors are shouting about bubbles while scrambling to get on board.

5. CoreWeave Beats Earnings Expectations: Q2 Revenue Doubles, Pre-Market Stock Up 18%

Summary: AI cloud giant CoreWeave reported Q2 earnings with doubled revenue, driven primarily by surging demand for AI compute. The report was praised as a "cleaner quarter," and pre-market stock price rose as much as 18%.
Source: CNBC · 2026-08-13
Editor's Take: Is selling compute a steadier business than selling chips? CoreWeave is the earliest sample of Nvidia's "financialization of compute" experiment.

6. Cisco Beats Earnings but Stock Falls: AI Network Demand Drives Orders

Summary: Cisco reported earnings where both revenue and guidance beat expectations, stating that demand for network equipment from AI data centers reached "broad and record" levels. However, the stock fell in after-hours trading as the market had divergent views on valuation.
Source: CNBC · 2026-08-13
Editor's Take: The "water sellers" of AI infrastructure aren't just Nvidia—switches, optical modules, and cooling systems are quietly making money. The lack of stock growth suggests good expectations have already been priced in.

7. Cerebras Stock Drops 14% After Earnings: Hardware Business Declines, Demand "Irregular"

Summary: AI chip firm Cerebras released its second post-IPO earnings report, raising full-year guidance, yet the stock still plummeted about 14%. Its hardware business declined, with demand described as "lumpy and irregular," and cloud business growth couldn't mask hardware shrinkage.
Source: CNBC · 2026-08-13
Editor's Take: The story of challenging Nvidia isn't easy to tell. Hardware sales are slumping, and the cloud business is still burning cash; the market votes with its feet.

8. Musk's SpaceXAI Launches Grok Bot: Enabling AI to Collaborate Like a Team

Summary: SpaceXAI launched Grok Bot, supporting multiple AI agents working together—a "captain" breaks down tasks, and multiple "team members" execute them in parallel, allowing users to assign work to AI like managing a team. Grok has upgraded from a chat assistant to an "AI teammate" capable of assigned work.
Source: Hacker News · 2026-08-13
Editor's Take: Lone-wolf chatbots are evolving into "AI work squads." This is the true beginning of Agents changing office dynamics.

III. Humanoid Robots

1. South Korea's Sovereign Wealth Fund Enters the Fray: Joining the Global AI and Robotics Race

Summary: South Korea's sovereign wealth fund (KIC) announced it will participate in the global AI and robotics investment race, planning to allocate funds to support domestic AI infrastructure and the robotics industry. Previously, South Korea set a goal to become one of the top three global robotics powers by 2030.
Source: Bloomberg · 2026-08-13
Editor's Take: National teams entering the robot buying game signals that humanoid robots have upgraded from "tech toys" to "national-level industries." Countries are all vying for the position of the next automotive industry.

2. Uber Dumps All Shares in Serve Robotics

Summary: Uber sold its entire stake in autonomous delivery robot company Serve Robotics. Serve was spun off from Uber five years ago as an unmanned delivery company; now the two have completely cut ties.
Source: Hacker News · 2026-08-13
Editor's Take: Unmanned delivery has retreated from "giant's innovation experiment" to "startup's solo act." In a period of capital ebb, surviving matters more than telling stories.

IV. Autonomous Driving

1. AI Agents Autonomously Launch Cyberattack: "Drill" Targeting Taiwan Sparks Debate

Summary: Research institutions demonstrated multiple AI agents autonomously completing a full cyberattack process without human intervention, targeting an environment in a Taiwanese institution. The demo sparked intense discussion on AI safety boundaries and autonomous weapons.
Source: Hacker News · 2026-08-13
Editor's Take: When attack chains can be fully automated, the defender's reaction time compresses from "hours" to "seconds." The AI offense-defense arms race has begun.

V. World Models / Physical AI

1. Amazon Uses Twitch Streamer Content to Train AI: Enabled by Default, Opt-Out Available

Summary: Twitch announced that streamers' live content will be used by default to train Amazon's generative AI models, with streamers able to manually opt out. The news triggered strong backlash in the creator community, dubbed "another non-consensual data harvest."
Source: TechCrunch · 2026-08-13
Editor's Take: Default on, opt-out later—the platform's consistent logic of "trading data for service." Content creators' labor is becoming free fuel for AI, and this contradiction will only sharpen.

2. Swiss Public Broadcaster Blocks AI Crawlers: Green Light Only for Domestic Swiss Models

Summary: Swiss public broadcasting group SRG announced it would block external AI crawlers from scraping its content, but made an exception for models developed domestically in Switzerland to protect the local media ecosystem and AI sovereignty.
Source: Hacker News · 2026-08-13
Editor's Take: Content providers are starting to counter US tech giants with the mindset of "my data feeds my AI." The battle for data sovereignty is spreading from the internet to every nation.

Track Statistics:

Track	Count
I. AI Large Models	6
II. AI Software	8
III. Humanoid Robots	2
IV. Autonomous Driving	1
V. World Models / Physical AI	2

Frontiers of the Physical World | WestTide · AI Industry Daily

Policy