MiniMax wants to find the next “10x”.
```
With the explosion of ClaudeCode, AI has shifted from a chat tool to an Agent. As models begin to truly perform tasks for people, Token consumption will see exponential growth.
Whoever enables AI to truly enter the production process will secure the most stable and sustainable Token consumption. After the collective surge and rapid pullback of AI concept stocks earlier this year, domestic large model companies are now seeking new narratives for growth.
Having capitalized on vibe coding and the lobster craze, and reaped the rewards, domestic model player MiniMax is eager to expand its circle of friends and search for the next gold mine.
On May 11, MiniMax launched a new partnership program called "10xTeam".
In addition to its already established focus in vertical fields such as industrial software, game engines, chip design, finance, and accounting, this time MiniMax is mainly inviting experts in economics, life sciences, and materials chemistry—fields that are globally oriented and could deeply integrate with large models—to co-create, and has simultaneously posted "10xTeam Researcher" positions on job platforms.
The ambition behind this is obvious: they want to replicate the "10x efficiency leap" seen in programming across more industries.
This would be a win-win, as MiniMax boosts its foundational general intelligence capabilities through these partnerships and drives its models to penetrate deeper into industrial scenarios.
In fact, "general large model + industry expert co-building" has become consensus among leading companies.
Anthropic has long recruited academic and industry researchers; its EconomicIndex brings models' impact on industry economic activities into its evaluation perspective; OpenAI has launched HealthBench for healthcare and prioritized legal and financial settings for GPT optimization; Google DeepMind champions "scientific breakthroughs" as its flag: AlphaFold (structural biology), GNoME (materials science), etc., proving that top experts joining forces with fundamental research teams can generate "domain-level leaps".
By the end of 2025, Baidu also launched a similar "Wenxin Mentor" program targeting industry and academic experts, guiding large models in knowledge transfer, quality assessment, and professional calibration.
In the past year, programming has become the earliest field where large models showed "10x efficiency": tools like Cursor, ClaudeCode have already reshaped software development workflows, and basic infrastructure competition is largely settled.
After ClaudeCode went viral, the AI industry quickly reached a consensus: AI's most important ability is no longer "answering questions", but "completing tasks". Once AI enters real production systems, it becomes essential.
Programmers call it every day, businesses run it daily, teams integrate it for continuous collaboration, and inference chains keep growing. Model invocation shifts from occasional needs to sustained consumption, and Token revenue naturally grows exponentially.
Such certainty also attracts players dividing the pie. Eighteen months ago, AI coding was dominated by Copilot. Today, overseas players like Cursor, Windsurf, Cline, Claude Code, Aider are fiercely competing, while domestically DeepSeek TUI, Kimi Code, MiniMax M2.5, ByteDance's Trae, Tongyi Lingma, Wenxin Kuaima, Zhipu's CodeGeeX, Alibaba's Qoder, and more are vying for market share.
As programming's dividends reach a bottleneck, "What is the next field to be 10x'd?" becomes the question all companies must answer.
MiniMax's answer is: bring model capabilities into fields with high density of expert knowledge, complex workflows, and where standardized approaches have yet to form.
This is precisely the problem that cannot be solved by model teams optimizing behind closed doors. Top domain experts must define problems, co-build evaluation and workflows, and then let the model drive industry transformation.
Industry knowledge has inherently high barriers.
Chip design involves complex verification processes; industrial software houses vast engineering systems; finance has risk control logic and regulatory frameworks; life sciences are imbued with tacit experimental experience and specialized knowledge structures. None of these are naturally present in public internet data.
An industrial Agent that is truly usable faces challenges not in its reasoning capabilities, but in understanding industry workflows.
This is making large model companies increasingly resemble hybrid organizations combining research institutes, industry bodies, and consultancy firms. MiniMax's "10xTeam" is, in some sense, the first explicit move by a domestic large model vendor to adopt this "scientific collaboration model".
In MiniMax's view, this is akin to an industrial research partner mechanism: model teams handle foundational capabilities, while industry experts define problems, construct workflows, and establish evaluation systems, then Agents enter practical production settings.
Because as AI's goal shifts from "answering questions" to "completing tasks", the importance of industry experts grows rapidly.
Looking back, the most important talent in the internet era was the product manager, as they defined user needs. In the Agent era, the truly important people might be those who best understand industrial processes.
Programming is merely the first industry to be restructured by Agents. All large model companies now want to find the next scenario with massive Token consumption and genuine industrial value creation.
In the past year, the speed of large model industry valuation growth has evoked parallels to the internet bubble around 2000.
Recently, economist Ma Guangyuan pointed out: upstream infrastructure—computing power, optical modules, hardware—does have orders, revenue, and profitability, as global demand for compute is surging; but midstream large models and downstream applications like humanoid robots, general AI, and ToC/ToB implementation remain concepts and stories, without large-scale commercial use, sustained profits, or breakout real demand—yet all these future expectations are baked into current valuations.
The whole industry knows: if AI can't truly enter industries and help enterprises sustainably make money and improve efficiency, this capital game cannot last long. Only when AI starts working for enterprises, participating in production, and helping industries profit, can the entire industrial chain start running.
This is why leading AI companies worldwide are pushing into industrial deep waters.
Anthropic now emphasizes not just model ability but how Claude enters enterprise workflows; OpenAI strengthens vertical scenarios in healthcare, law, and finance; Google DeepMind maintains "scientific breakthroughs" as a strategic pillar.
Because everyone knows, AI must actually help industries make money, improve efficiency, and reduce costs for the industrial narrative to move forward. Otherwise, the bubble will inevitably burst.
And when the bubble bursts, it won't just affect a few model companies. From GPUs to cloud vendors, from data centers to AI startups, from primary to secondary markets, the entire AI value chain may endure a harsh winter.
So all large model companies today are racing against time to prove one thing: AI is not just a concept, but real productivity. MiniMax's "10xTeam" is essentially an industrial positioning move against this backdrop.
They hope to bind domain experts in advance, embed model capabilities into chip design, industrial software, financial analysis, life sciences and other complex industrial workflows, gradually building their own data, workflow, and commercialization moats.
Because as AI's goal shifts from "answering questions" to "completing tasks", industry knowledge becomes the new scarce resource. Programming is only the first industry to be restructured by Agents. What the whole AI sector truly wants to prove now: Will the next be the entire industrial world?
Risk Warning and DisclaimerThe market carries risks and investments require caution. This article does not constitute personal investment advice, nor does it take into account individual users' specific investment targets, financial situation or needs. Users should consider if any opinions, views or conclusions in this article fit their own circumstances. Investing accordingly is at your own risk. ```