Jensen Huang: The collaboration period with SK Hynix will exceed two years 黄仁勋：与SK海力士的合作期限将超过两年

Shifting focus from Silicon Valley’s chip aristocracy to the domestic scene, a company called “HuaChao ShenKong” just secured a hundred-million-dollar angel round, with a name straight out of a martial arts manual—“AI Ultrasound Brain-Machine.” Jingwei, Delian, Daoyuan, and a host of well-known VCs are racing to invest. In the brain-computer interface field, Elon Musk’s Neuralink is putting monkeys on the altar of video games, while countless pipelines in global labs are still stuck in papers and prototype stages. Demanding a “hundred-million-dollar” valuation at the angel round—this bubble is more elusive than the brain signals they aim to decode. Technical validation and industrialization? The road is long enough to circle the earth three times. Are the investors’ generosity today a bet on the stars and seas a decade away, or on a media headline for cashing out next quarter? In this field, the flashy demo you see today is likely its highest point ever.

Placing a few trending news stories together, the concentration of magical realism is off the charts. “After using AI, the company seems poorer”—this headline is practically a wail. How many bosses jumped in with heavy investment chasing the fairy tale of “cost reduction and efficiency improvement,” only to end up with unusable models, unaffordable API bills, and more anxious employees? AI hasn’t become the engine of profit but first became a bottomless pit of costs. Next up: “A trillion-dollar AI company bans using AI during interviews.” See, the shovel-selling giants themselves know how unreliable this tool is in the “craftsman” domain of creating real value. They use AI to optimize resume screening and interview scheduling, but absolutely prohibit candidates from cheating when showcasing their true skills. This is the ultimate performance art: while hyping AI as a disruptor of all jobs, they fear AI’s distortion in the most critical talent assessment. It’s as if they’re saying: you can play with AI, but for real skill, show those hands untainted by algorithms.

Then there’s “OpenAI chip veteran joins Anthropic”—does it resemble martial arts masters drifting between major sects? In the ultimate competition for AGI (Artificial General Intelligence), the movement of top “weaponsmiths” (chip experts) speaks louder than any financial report about shifts in power. The faster OpenAI’s commercialization wagon races, the more likely its research idealism gets left behind. Talents vote with their feet. Their migration paths quietly sketch a draft of the future AI power map.

The entire AI industry now resembles a massive overheating engine, fueled by capital and anxiety, with data and computing power burning in the cylinders. The Jensens of the world count money at the top of the industrial chain, the HuaChao ShenKongs sing their triumph in financing news, while countless small and medium-sized companies grow “poorer the more they use AI” through trial and error, and giants quietly erect “No AI Allowed” signs in their backyards. Mania and sobriety, triumphs and laments, promises and precautions—all coexist in the same time and space in a fractured manner. We’ve had enough grand narratives about changing the world; it’s time to sniff out the overly real, burnt-money, and humanly complex undertone beneath. When the tide recedes, who’s skinny-dipping remains unknown, but one thing is certain: those who first cry cold and poor are always the ones washed ashore by the waves, yet think they’ve grasped a pearl.

黄仁勋和SK海力士这对“芯片鸳鸯”又要续签婚约了，说是合作期超过两年，还留了“不断延长”的余地。这话听着像蜜糖，但嚼起来全是金属味。英伟达每年从SK海力士那儿搬走数十亿美元的HBM高带宽内存，今年采购额还要“大幅增长”。这哪是合作，分明是AI算力军备竞赛下，一个手握王牌的军火商对关键零部件供应商的深度绑定与战略锁喉。黄仁勋的轻描淡写背后，是HBM成了比石油还紧俏的硬通货，是SK海力士的工厂可能比印钞厂还忙。这场婚姻里，甜蜜是暂时的，权力不对等才是常态。当全世界的AI公司都排着队从黄仁勋手里买铲子，他就成了那个决定谁先挖到金矿的人，而SK海力士，成了那把铲子最关键的握柄。

视线从硅谷的芯片豪门转会到国内，一家叫“华超神控”的公司刚揣着亿元级的天使轮融资，名字起得跟武侠秘籍似的——“AI超声脑机”。经纬、德联、道远，一票知名VC抢着送钱。脑机接口这领域，一边是马斯克的Neuralink把猴子送上电子游戏的圣坛，一边是全球实验室里无数的管线还卡在论文和原型机阶段。天使轮就敢要“亿元级”，这估值泡沫吹得比他们要解读的脑电信号还飘忽。技术验证和产业化落地？路还长得能绕地球三圈。投资人们此刻的慷慨，赌的究竟是十年后的星辰大海，还是下一个季度就套现离场的媒体头条？在这个领域，今天你看到的炫酷demo，很可能就是它这辈子最高光的时刻。

把几条热榜新闻摆在一起看，魔幻现实主义浓度直接超标。“用了AI之后，公司好像更穷了”，这标题简直是一声悲鸣。多少老板冲着“降本增效”的童话重金入局，结果买了一堆用不起来的模型、付不起的API账单，以及更焦虑的员工。AI没成为利润的发动机，先成了成本的无底洞。紧接着另一条：“万亿美元AI公司，面试时禁止使用AI”。看吧，卖铲子的巨头自己都清楚，这工具在创造真实价值的“手艺人”领域有多不可靠。他们用AI优化简历筛选、面试安排，却绝对不容许候选人在展示真实技艺时作弊。这简直是最大的行为艺术：一边鼓吹AI将颠覆一切工作，一边在最关键的人才鉴别环节，恐惧AI带来的失真。仿佛在说，你们可以用AI玩闹，但真本事，请亮出那双没有受算法沾染的、原生态的手。

再看那条“OpenAI芯片元老加入Anthropic”，像不像武林高手在各大门派间的漂移？在AGI（通用人工智能）这场终极争霸赛里，顶尖的“武器匠”（芯片专家）的流向，比任何财报都能说明势力的消长。OpenAI的商业化马车跑得越快，研究上的理想主义就越可能被甩下车。人才，是会用脚投票的。他们的迁徙路径，悄悄画出了一张未来AI权力地图的草图。

整个AI行业现在就像一个巨大的过热发动机，燃料是资本和焦虑，汽缸里燃烧的是数据和算力。黄仁勋们站在产业链顶端数钱，华超神控们在融资新闻里高歌猛进，而无数中小公司在试错中“越用越穷”，巨头们则在自家后院悄悄立起“此地禁用AI”的牌子。狂热与冷静，高歌与哀鸣，承诺与戒备，在同一个时空里撕裂式共存。我们听够了改变世界的宏大叙事，是时候该闻一闻这宏大叙事下，那些过于真实的、带着烧钱焦糊味和人性复杂气息的底噪了。当潮水退去，谁在裸泳尚未可知，但可以肯定的是，最先喊冷、喊穷的，永远是那些被浪花拍在沙滩上，却以为自己抓到了珍珠的人。