在 2 月 2 日前约 4 年，颜俊杰（Yan Junjie）把 MiniMax 推介为一家多模态 AI 新创（文字、图像、音讯、影片），却广泛被斥为骗子；到 2 月 2 日，这家总部位于上海的公司被形容为数十亿美元级，颜也在 36 岁成为亿万富翁。MiniMax 的模型被执著地拿来与 OpenAI 做基准比较，并被置于更广泛的、由中国推动的技术自主浪潮之中，背景是日益升级的美中裂痕。文章对比早期高能见度的科技名人时代，例如马云（Jack Ma）在 2017 年阿里巴巴周年庆上的表演，与新一代更低调的创办人群体：他们在打造 AI 基础设施、软体，以及机器人/电动车（EV）的「AI 机械」。

关键的财富与规模数字构成主轴：一批新的、与中国 AI 相关的富豪据称合计拥有 $100.5 billion，接近 Bill Gates 的 $105 billion，但远低于 Jensen Huang 的 $153 billion。个别例子包括寒武纪（Cambricon）的陈天石（Chen Tianshi），其财富自 2024 年初以来上升逾 800% 至 $21.5 billion，同时 2025 年营收增幅逾 600%；摩尔线程（Moore Threads）创办人张建中（Zhang Jianzhong）用 5 年把公司做成 $45 billion；沐曦（MetaX）创办人陈维良（Chen Weiliang）的 IPO 据称首日上涨 692%；DeepSeek 的梁文峰（Liang Wenfeng）被认为推出了 R1 模型，据报训练成本为 $5.5 million，而可比的美国投入常被引述为 $100 million+。文章也把成长与政策及制裁连结起来：寒武纪在 12/2022 被列入美国 Entity List，摩尔线程在 10/2023，Intellifusion 在 06/2020；国内替代压力包括对敏感领域提出的 100% 本地晶片采购要求，以及国家资本例如向长鑫存储（ChangXin Memory Technologies）注入 $2 billion 的「Big Fund」。

其影响强调：中国在科技领域的财富创造日益与国家产业政策绑定，并受制裁风险塑形，促使创办人刻意保持匿名，以避免美国限制与国内审视；该群体被描述为几乎清一色男性，出生于 1970s 与 1980s，且多具精英学术背景（例如 16 岁入大学、Ph.D. 资历、以及长期在国家研究体系任职）。叙事不仅止于软体，也延伸到「picks and shovels」式的基础设施（晶片、电路板、光收发器）与机器人，并举例里程碑，如 1,300 台机器人同时跳舞的 Guinness 纪录，以及地方政府的企业注册在 4 小时内获批。方法论的保留亦很明确：身家与估值（尤其私营公司）依赖同业比较、融资轮次与申报文件；数字截至 2 月 2 日；像 Alibaba 与 Tencent 这类多元化巨头被排除；而梁文峰的身家被标示为高度不确定，范围约从 $1 billion 到超过 $150 billion。

About 4 years before Feb. 2, Yan Junjie pitched MiniMax as a multi-modal AI startup (text, images, audio, video) and was widely dismissed as a fraud; by Feb. 2, the Shanghai-based company is described as multibillion-dollar, and Yan is a billionaire at age 36. MiniMax’s models are benchmarked obsessively against OpenAI and are framed as part of a broader China-driven push for technological independence amid an escalating US-China rift. The article contrasts an earlier era of high-visibility tech celebrities, exemplified by Jack Ma’s 2017 Alibaba anniversary performance, with a newer, quieter cohort of founders building AI infrastructure, software, and robotics/EV “AI machinery.”

Key wealth and scale figures anchor the trend: a set of new Chinese AI-linked tycoons is said to have a collective $100.5 billion, near Bill Gates’s $105 billion but far below Jensen Huang’s $153 billion. Individual examples include Cambricon’s Chen Tianshi, whose wealth is up more than 800% since the start of 2024 to $21.5 billion, alongside a revenue surge of more than 600% in 2025; Moore Threads founder Zhang Jianzhong built it over 5 years into a $45 billion company; MetaX founder Chen Weiliang’s IPO allegedly jumped 692% on day 1; DeepSeek’s Liang Wenfeng is credited with an R1 model reportedly trained for $5.5 million versus the $100 million+ often cited for comparable US efforts. The article also ties growth to policy and sanctions: Cambricon was added to the US Entity List in 12/2022, Moore Threads in 10/2023, and Intellifusion in 06/2020, while domestic substitution pressures include a stated 100% local-chip sourcing requirement for sensitive sectors and state capital like a $2 billion “Big Fund” injection into ChangXin Memory Technologies.

The implications emphasize how China’s wealth creation in tech is increasingly coupled to state industrial policy and shaped by sanctions risk, pushing founders toward deliberate anonymity to avoid US restrictions and domestic scrutiny; the cohort is described as overwhelmingly male and born in the 1970s and 1980s, often with elite academic backgrounds (for example, entry to college at 16, Ph.D. credentials, and long stints in state research). The narrative extends beyond software to “picks and shovels” infrastructure (chips, circuit boards, optical transceivers) and robotics, citing milestones like a Guinness record of 1,300 robots dancing simultaneously and a local-government business registration approved in 4 hours. Methodology caveats are explicit: net worth and valuation estimates (especially for private firms) rely on peer comparisons, funding rounds, and filings; numbers are as of Feb. 2; diversified conglomerates like Alibaba and Tencent are excluded; and Liang Wenfeng’s net worth is flagged as highly uncertain, spanning roughly $1 billion to more than $150 billion.