MiniMax Takes a Shot at OpenAI and…

1. Key Themes

Theme: The frontier-model race is intensifying on price/performance and IPO positioning

Chinese labs are attacking the leaders on cost and speed

MiniMax claims parity or better against the top US models at far lower cost.

"Chinese AI startup MiniMax just released M3.1 Flash, an update it says runs nearly three times faster and costs 70% less to use. By MiniMax's own benchmarks, the model beats OpenAI's GPT-5.6 Sol and Anthropic's Claude Opus 5 on some agentic tasks and performs about the same on coding."

OpenAI is front-running Anthropic's IPO to control the valuation narrative

"OpenAI is racing to close a roughly $30 billion round at a valuation above $1.4 trillion before Anthropic's expected mid-November IPO, seeking to lock up overlapping investors and establish its valuation before Anthropic's disclosures invite direct comparisons."

Public-market investors may get weak governance rights in Anthropic

"Anthropic's planned IPO will leave public investors with unusually limited governance rights because its opaque Long-Term Benefit Trust can appoint most of the company's board and block major corporate moves, with key details about the trust's powers and governing documents still undisclosed."

Theme: AI infrastructure is raising mega-rounds, but public-market appetite is cooling

Private capital is still flowing to AI compute alternatives

"Tenstorrent, a 10-year-old Toronto company that builds RISC-V-based AI processors, servers, workstations, and open-source software, raised a $1.5 billion Series E round at a $6.2 billion post-money valuation."

Public investors are pushing back on AI infrastructure valuations

"Nscale is facing investor pushback over its planned IPO valuation, with some prospective buyers urging the Nvidia-backed AI data-center developer to target $16 billion to $19 billion rather than the roughly $30 billion it has been seeking amid broader pressure on AI infrastructure stocks."

New funds are being raised around AI adoption and compute

"Inovia Capital... has raised a $200 million first close for its sixth venture fund, which will target 25 to 30 startups in sectors where AI adoption is still early but poised to have a meaningful economic impact."

"Lux Capital... disclosed plans to raise up to $300 million for Lux Compute Leaders, a new investment vehicle..."

Theme: AI security and geopolitics are becoming operational risks

Distillation and API-key abuse create real costs for developers

"OpenCode... said Saturday that attackers broke into an older system overnight, affecting 1.4% of users of its online platform, OpenCode Console, where some users store the API keys, or digital passwords, they use to access AI services."

"Other users reported charges in the hundreds of dollars on individual accounts; a few described totals above $1,000."

Defenders lack access to less-restricted models

"Companies without vetted access to AI models with fewer safety restrictions have a hard time defending against large-scale, AI-driven cyberattacks."

Government action against Chinese AI vendors is escalating

"The Department of Homeland Security ordered federal civilian agencies to remove DeepSeek technology from their IT systems and stop buying its products. The ban also covers offerings from DeepSeek's affiliates and any third-party services built on DeepSeek."

Insider and espionage risk reaches AI labs

"Federal prosecutors say an OpenAI engineer charged with planting a recording device in an executive conference room was in frequent contact with a Saudi royal who sought to recruit him."

Theme: Capital is concentrating in AI agents/automation and in physical-world AI

Agent and assistant startups are commanding unicorn-level valuations

"Town, a two-year-old San Francisco startup that provides AI assistants that coordinate work across teams and communicate with customers, partners, and candidates, raised a $108 million Series B round at a $1 billion post-money valuation."

Robotics foundation models are getting huge early checks

"Dream Labs, a San Francisco startup founded this year that builds a foundation model to train robots from video and natural-language instructions, raised a $90 million seed round..."

Vertical AI agents are attracting early-stage money

"Vessla, a Stockholm startup founded this year that uses AI agents to automate workflows across existing software for residential and commercial property operators, raised a $3.3 million pre-seed round..."

2. Contrarian Perspectives

Skeptics doubt the "distillation attack" framing of the OpenCode breach

The company's framing implies a sophisticated, AI-relevant attack, but some dispute it.

"Some skeptics called the framing an overstretch, arguing that the incident was a straightforward breach whose use for distillation remains unproven."

The company itself "didn't specify how the distillation attack occurred," and notes only that "leaked API keys can help users in countries with restricted access to American models generate the high volume of outputs needed for that process." The operational lesson (credential hygiene) holds either way, but the narrative may be inflated.

Anthropic's mission-driven governance may be a liability for public investors

Anthropic's safety-oriented structure is usually viewed as a feature; the IPO reframes it as an investor-rights problem.

"Public investors [will have] unusually limited governance rights because its opaque Long-Term Benefit Trust can appoint most of the company's board and block major corporate moves, with key details about the trust's powers and governing documents still undisclosed."

The undisclosed trust details are the sticking point, and they could affect pricing as comparisons with OpenAI sharpen.

AI infrastructure private valuations may not survive public scrutiny

Private markets are pricing AI compute aggressively (Tenstorrent at $6.2B post-money after a $1.5B round), but public-market buyers are signaling a haircut.

"...some prospective buyers urging the Nvidia-backed AI data-center developer to target $16 billion to $19 billion rather than the roughly $30 billion it has been seeking amid broader pressure on AI infrastructure stocks."

A gap of roughly 37–47% below the ask suggests the private-to-public valuation bridge is under strain.

3. Companies Identified

MiniMax

Description: Chinese AI/LLM startup.
Why mentioned: Released M3.1 Flash, claiming speed and cost advantages over US frontier models.
Quote: "an update it says runs nearly three times faster and costs 70% less to use."

OpenAI

Description: Frontier AI lab.
Why mentioned: Pursuing a ~$30B round at $1.4T+ valuation ahead of Anthropic's IPO; also named in an insider-threat case.
Quote: "OpenAI is racing to close a roughly $30 billion round at a valuation above $1.4 trillion before Anthropic's expected mid-November IPO..."

Anthropic

Description: Frontier AI lab preparing an IPO.
Why mentioned: Expected mid-November IPO; governance structure under scrutiny.
Quote: "Anthropic's planned IPO will leave public investors with unusually limited governance rights..."

DeepSeek

Description: Chinese AI company.
Why mentioned: Banned from US federal civilian agency systems.
Quote: "The ban also covers offerings from DeepSeek's affiliates and any third-party services built on DeepSeek."

OpenCode

Description: Open source AI coding agent.
Why mentioned: Security breach affecting users' API keys.
Quote: "an open source AI coding agent that has been gaining ground among developers as an alternative to Claude Code and Cursor"

Tenstorrent

Description: Toronto-based builder of RISC-V AI processors, servers, workstations, and open-source software.
Why mentioned: $1.5B Series E at $6.2B post-money.
Quote: "raised a $1.5 billion Series E round at a $6.2 billion post-money valuation."

Nscale

Description: Nvidia-backed AI data-center developer.
Why mentioned: Pushback on its IPO valuation.
Quote: "...target $16 billion to $19 billion rather than the roughly $30 billion it has been seeking..."

Town

Description: San Francisco AI assistants for cross-team and external coordination.
Why mentioned: $108M Series B at a $1B valuation.
Quote: "raised a $108 million Series B round at a $1 billion post-money valuation."

Dream Labs

Description: Foundation model for training robots from video and language.
Why mentioned: $90M seed round.
Quote: "raised a $90 million seed round co-led by B Capital and Kindred Ventures"

Abby Care

Description: Trains and pays family members to provide Medicaid-covered home care.
Why mentioned: $82M Series C.
Quote: "raised an $82 million Series C round led by Transformation Capital"

Even Healthcare

Description: Bengaluru hospital operator and preventive care coordinator.
Why mentioned: $50M Series B at $300M post-money.
Quote: "raised a $50 million Series B round at a $300 million post-money valuation."

Nouscom

Description: Basel-based immunotherapy developer.
Why mentioned: $134.9M Series D.
Quote: "develops immunotherapies that train the immune system to recognize and attack cancerous and precancerous cells"

Prism Data

Description: Bank-transaction analysis for lender credit assessment.
Why mentioned: $20M Series A.
Quote: "analyzes bank-account transactions to help lenders assess creditworthiness"

Tolan

Description: Voice-first AI companion with long-term memory.
Why mentioned: $30M round.
Quote: "has raised a total of $60 million."

Kodama

Description: Remote-operation and supervised-autonomy systems for forestry equipment.
Why mentioned: $7M round.
Quote: "builds remote-operation and supervised-autonomy systems for forestry equipment"

Linkener

Description: Valencia-based energy monitoring software and metering hardware.
Why mentioned: $8.4M round.
Quote: "help businesses track consumption, validate bills, and analyze electricity contracts"

Vessla

Description: Stockholm AI-agent workflow automation for property operators.
Why mentioned: $3.3M pre-seed.
Quote: "uses AI agents to automate workflows across existing software for residential and commercial property operators"

Inovia Capital

Description: Montreal VC firm.
Why mentioned: $200M first close on Fund VI targeting early-AI-adoption sectors.
Quote: "target 25 to 30 startups in sectors where AI adoption is still early but poised to have a meaningful economic impact."

Lux Capital

Description: New York VC firm.
Why mentioned: Raising up to $300M for Lux Compute Leaders.
Quote: "disclosed plans to raise up to $300 million for Lux Compute Leaders"

Permanent Partners Collective

Description: New VC fund focused on AI applications and infrastructure.
Why mentioned: Targeting $100M+.
Quote: "targeting more than $100 million for Permanent Partners Collective Fund I"

Google (Gemini 4 Argon)

Description: Google's model family.
Why mentioned: Multi-agent system beat Minecraft autonomously.
Quote: "Four AI agents running on Google's Gemini 4 Argon beat Minecraft without human help, autonomously dividing tasks, sharing warnings and memories, and writing 2,783 JavaScript programs over 27 hours before defeating the Ender Dragon."

iLands

Description: Social media app for AI agents.
Why mentioned: Agents autonomously cold-email humans to earn operating tokens.
Quote: "AI agents on the social media app iLands are autonomously cold-emailing writers, researchers, and literary magazines for paid work and submissions to earn the tokens they need to keep operating."

Hart (bakery chain)

Description: Copenhagen bakery chain.
Why mentioned: Sells 50,000 cheese buns a month.
Quote: "sells 50,000 a month at bakery chain Hart."

4. People Identified

Eric Schmidt

Description: Former Google CEO and early Anthropic investor.
Why mentioned: Positioned for a major windfall from the IPO.
Quote: "owns roughly 1% of Anthropic after investing tens of millions of dollars in the AI startup's Series A, potentially setting him up for one of the largest individual windfalls from its imminent IPO."

Elon Musk

Description: SpaceX owner/CEO.
Why mentioned: Plans to give SpaceX shares to every US child.
Quote: "plans to give one of his SpaceX shares to every child in the U.S. in 2027, a giveaway worth roughly $10.5 billion at Friday's closing price"

Ayesha Arora-Hughes

Description: Former S32 investor, new fund founder.
Why mentioned: Raising a $100M+ AI-focused fund, advised by Palo Alto Networks CEO Nikesh Arora.
Quote: "targeting more than $100 million for Permanent Partners Collective Fund I, a new VC fund focused on AI applications and infrastructure, with her father, Palo Alto Networks CEO Nikesh Arora, serving as an adviser."

Nathan Fielder

Description: Comedian/filmmaker.
Why mentioned: Profiled over his documentary on Elizabeth Holmes.
Quote: "Holmes unexpectedly calling him from prison days before release to tell him twice, 'I trust you.'"

Elizabeth Holmes

Description: Theranos founder, subject of Fielder's work.
Why mentioned: Called Fielder from prison before release.
Quote: "an open-ended attempt to understand Elizabeth Holmes"

5. Operating Insights

Treat API keys as high-risk liabilities

The OpenCode incident shows stored keys can trigger real financial damage across multiple providers.

"I saw my Open AI, Anthropic, and Google AI spend alerts all fire off before going to bed wondering what was going on."

Set spend alerts and hard caps on every provider, and avoid storing long-lived keys on third-party platforms.

Audit vendor exposure to Chinese AI models if you sell to government

The DHS ban reaches beyond DeepSeek itself.

"The ban also covers offerings from DeepSeek's affiliates and any third-party services built on DeepSeek."

Map your dependency chain, since downstream products built on a banned model can be swept in.

Design agent systems with coordination and shared memory

The Minecraft result points to an architecture pattern: specialization plus communication.

"...autonomously dividing tasks, sharing warnings and memories, and writing 2,783 JavaScript programs over 27 hours before defeating the Ender Dragon."

For long-horizon workflows, multi-agent role division with shared state may outperform single-agent approaches.

6. Overlooked Insights

Agents are starting to earn their own compute

The iLands detail signals an early version of autonomous economic actors.

"AI agents... are autonomously cold-emailing writers, researchers, and literary magazines for paid work and submissions to earn the tokens they need to keep operating."

If agents self-fund inference, it raises questions about spam, payments infrastructure, and who bears liability for their outreach.

Women use and trust AI less, and men overclaim AI knowledge

This may matter for product adoption, trust design, and market sizing.

"A Harvard survey of 5,403 U.S. adults found that women use and trust AI less than men do... men were more likely than women to claim they knew every name on the list, fakes included."

Products that rely on self-reported familiarity or confidence could skew toward overconfident users, and an undertapped segment may respond to stronger trust signals.