首页 500强 活动 榜单 商业 科技 商潮 专题 品牌中心
杂志订阅

独家:密歇根出身的特斯拉前高管完成4500万美元融资

Nick Lichtenberg
2026-10-21

从当前阶段走向真正工业化量产的这一步,才是最大难关。

文本设置
小号
默认
大号
Plus(0条)

Holyvolt首席运营官史蒂文·普拉斯特。图片来源:courtesy of Holyvolt

史蒂文・普拉斯特年纪够大,还记得手机需放进包里携带的年代。那时手机太重,没法随身揣着,多数人都将其装在车上。在接受《财富》Zoom 视频采访时,普拉斯特说,最终让手机缩小到可以戴在手腕上的,与其说是晶体管,不如说是电池,他还举起自己的手机作为佐证。

“促成之一转变的其实是电池,”普拉斯特说。他曾任特斯拉制造业务负责人,现在是电池材料初创公司Holyvolt的首席运营官。

据他估算,锂离子技术在成熟曲线上最多刚走完两成,保守估计可能仅一成。他说,西方若想赢下后面那一大段,就不能只靠在格局已定、中国已占主导地位的成熟环节与后者竞争。

“在中国人自己定规矩的战场上,我们不可能比中国还中国,”普拉斯特说,“论做中国这件事,他们是世界冠军,无人能出其右。”

Holyvolt拿出了4500万美元布局替代路线,全部为股权融资。这家总部位于斯德哥尔摩的初创公司向《财富》透露,本轮融资由Fifth Wall、Zone II、CC.VC以及来自国防和金融行业的数位天使投资人参投。知情人士透露,本轮融资对Holyvolt的估值达2.45亿美元。

Holyvolt试图跨越的鸿沟极为巨大。中国生产全球约85%的正极活性材料(该材料很大程度上决定了电池的储能容量和成本)和全球约80%的电芯。中国的领先地位,源自将镍锰钴三元锂(NMC)和磷酸铁锂(LFP)这两种当下主流的化学体系,推到了无人能及的量产规模。

“如果我们选择磷酸铁锂路线,试图在这条路线上战胜中国,那我们必输无疑,”普拉斯特说,“而且是比一次输一次。”

他在一份声明中表示,出路在于“西方发明更好的材料和工艺,并在西方制造出来。”他告诉《财富》,他认为人类“还有十几种正极材料有待发现。”

一个化学问题

既然电池技术还远未成熟,为什么没能像芯片那样走出一条稳定、可预测的发展曲线?普拉斯特的解释是:摩尔定律源于物理学,人类对这门学科的理解足够深入,可以在纸面上推演求解,而电池属于另一门完全不同的学问。

“如果一架飞机是靠计算机计算设计出来的,那我敢坐,” 他说,“但化学里,还有太多未解之谜。”

“电池的麻烦在于,它是个化学问题,不是物理问题,” 他说,“要是物理问题,早就有摩尔定律了。”

在AI赋能科学研究这件事上,业界存在分歧。普拉斯特说,业内多数人认为,化学问题可以靠计算解决,少数人则认为,AI仍需依托实体实验室来验证其结果——Holyvolt就属于后者。

“我们得去实验室,得做实验。”普拉斯特说。

赌注

Holyvolt旗下实验室位于圣地亚哥,名为Wildcat Discovery Technologies(今年3月以7300万美元的价格与Holyvolt完成合并,Holyvolt也确认了这一数字)。Wildcat由斯克里普斯研究所化学家彼得・舒尔茨创立,自建有一套高通量系统,能并行测试数千种电池材料。该公司称,18年来,该平台已为包括宝马在内的超过85家客户完成超过225个项目,测试电芯超过50万枚。

这一大批数据,正是Holyvolt瑞典籍创始人兼首席执行官马蒂亚斯·英格瓦松创办这家公司时所依托的核心资产。英格瓦松曾是泰普尔(Tempur-Pedic)国际化扩张的核心推手之一,也对该公司在纽交所IPO前后取得的一系列成功做出过重大贡献,上市后还曾执掌该公司欧洲业务六年。他告诉《财富》,合并成一家公司后,Holyvolt已将Wildcat的实验结果用于AI模型训练,还将其反向接入实验室,让每一轮实测都能对下一轮预测进行优化。这些数据与实验室共同支撑着AI模型,让它既能满足市场需求,提前预判哪种新材料可能可行,也能加速研发,推动下一代突破性化学体系的开发工作。

多数“AI赋能科学”领域的初创企业都是先搭建模型,再搜集数据。英格瓦松在接受《财富》采访时称,Holyvolt走的是相反路径:“我们深耕强化学习20年,多年积累的研究成果,正是AI开展学习、迭代演进时可以依托的根基。”

“我们选择的是另一条技术路线,搭建这套体系难度极高,”他说,“需要经年累月的沉淀。”

英格瓦松表示,一项与客户共同完成的案例研究显示,该实验室的研发速度已达传统研发模式的10倍之多,而Holyvolt认为,一旦依托AI实现闭环迭代,相较于传统研发模式,这一速度还能继续提升。“我甚至可以说,我们已经找到了破解之道。”他说。

最后期限

该研发体系产出的首个成果是无序岩盐材料(disordered rocksalt,简称DRX),一类不含镍、不含钴的正极材料。普拉斯特称,其性能可媲美镍锰钴三元锂,“优于或等同于”镍基化学体系,在北美的落地成本与磷酸铁锂大致相当。他表示,无序岩盐的核心设计要求之一,是打造“完全安全可控”的供应链,但拒绝透露供应商信息。无序岩盐的性能数据已在2026年3月的国际电池大会(International Battery Conference)等多场学术会议上公布。

这一时间节点暗含一项硬性期限。2027年10月1日,美国《2024财年国防授权法》第154条将正式实施。该条款禁止美国国防部采购宁德时代、比亚迪以及另外四家列明的中国厂商生产的电池,而美军当下正着力扩大无人机的应用规模。普拉斯特称,Holyvolt计划在2027年实现无序岩盐材料的商业化落地,优先面向无人机领域,之后再拓展至交通出行市场。该公司同时在开发面向车企的电池配方,还在研发适配数据中心的专用方案(数据中心的GPU启动时,对瞬时峰值功率有极高要求)。

“这一幕我早见过”

对普拉斯特来说,这场较量有着特殊的个人意义。“我在密歇根北部长大,家族世代深耕汽车行业,”他说,“这一幕我早见过,结局如何,我心里有数。”

在他看来,这场较量更多关乎稳定,而非对抗竞争。“无论哪个领域,若资源过度集中于一处,局面就会失衡。” 他说。

英格瓦松也持类似观点。他说:“我对中国没有任何意见” ,并随即提到,他祖国最知名的车企沃尔沃,“如今已是中方旗下企业。”(中资车企吉利持有该公司多数股权。)

Holyvolt无意自建超级工厂。普拉斯特说,公司愿意开展技术授权、委托代工或合资合作,把工业化量产这部分业务留给合作伙伴。

从当前阶段走向真正工业化量产的这一步,才是最大难关。(财富中文网)

译者:梁宇

审校:夏林

史蒂文・普拉斯特年纪够大,还记得手机需放进包里携带的年代。那时手机太重,没法随身揣着,多数人都将其装在车上。在接受《财富》Zoom 视频采访时,普拉斯特说,最终让手机缩小到可以戴在手腕上的,与其说是晶体管,不如说是电池,他还举起自己的手机作为佐证。

“促成之一转变的其实是电池,”普拉斯特说。他曾任特斯拉制造业务负责人,现在是电池材料初创公司Holyvolt的首席运营官。

据他估算,锂离子技术在成熟曲线上最多刚走完两成,保守估计可能仅一成。他说,西方若想赢下后面那一大段,就不能只靠在格局已定、中国已占主导地位的成熟环节与后者竞争。

“在中国人自己定规矩的战场上,我们不可能比中国还中国,”普拉斯特说,“论做中国这件事,他们是世界冠军,无人能出其右。”

Holyvolt拿出了4500万美元布局替代路线,全部为股权融资。这家总部位于斯德哥尔摩的初创公司向《财富》透露,本轮融资由Fifth Wall、Zone II、CC.VC以及来自国防和金融行业的数位天使投资人参投。知情人士透露,本轮融资对Holyvolt的估值达2.45亿美元。

Holyvolt试图跨越的鸿沟极为巨大。中国生产全球约85%的正极活性材料(该材料很大程度上决定了电池的储能容量和成本)和全球约80%的电芯。中国的领先地位,源自将镍锰钴三元锂(NMC)和磷酸铁锂(LFP)这两种当下主流的化学体系,推到了无人能及的量产规模。

“如果我们选择磷酸铁锂路线,试图在这条路线上战胜中国,那我们必输无疑,”普拉斯特说,“而且是比一次输一次。”

他在一份声明中表示,出路在于“西方发明更好的材料和工艺,并在西方制造出来。”他告诉《财富》,他认为人类“还有十几种正极材料有待发现。”

一个化学问题

既然电池技术还远未成熟,为什么没能像芯片那样走出一条稳定、可预测的发展曲线?普拉斯特的解释是:摩尔定律源于物理学,人类对这门学科的理解足够深入,可以在纸面上推演求解,而电池属于另一门完全不同的学问。

“如果一架飞机是靠计算机计算设计出来的,那我敢坐,” 他说,“但化学里,还有太多未解之谜。”

“电池的麻烦在于,它是个化学问题,不是物理问题,” 他说,“要是物理问题,早就有摩尔定律了。”

在AI赋能科学研究这件事上,业界存在分歧。普拉斯特说,业内多数人认为,化学问题可以靠计算解决,少数人则认为,AI仍需依托实体实验室来验证其结果——Holyvolt就属于后者。

“我们得去实验室,得做实验。”普拉斯特说。

赌注

Holyvolt旗下实验室位于圣地亚哥,名为Wildcat Discovery Technologies(今年3月以7300万美元的价格与Holyvolt完成合并,Holyvolt也确认了这一数字)。Wildcat由斯克里普斯研究所化学家彼得・舒尔茨创立,自建有一套高通量系统,能并行测试数千种电池材料。该公司称,18年来,该平台已为包括宝马在内的超过85家客户完成超过225个项目,测试电芯超过50万枚。

这一大批数据,正是Holyvolt瑞典籍创始人兼首席执行官马蒂亚斯·英格瓦松创办这家公司时所依托的核心资产。英格瓦松曾是泰普尔(Tempur-Pedic)国际化扩张的核心推手之一,也对该公司在纽交所IPO前后取得的一系列成功做出过重大贡献,上市后还曾执掌该公司欧洲业务六年。他告诉《财富》,合并成一家公司后,Holyvolt已将Wildcat的实验结果用于AI模型训练,还将其反向接入实验室,让每一轮实测都能对下一轮预测进行优化。这些数据与实验室共同支撑着AI模型,让它既能满足市场需求,提前预判哪种新材料可能可行,也能加速研发,推动下一代突破性化学体系的开发工作。

多数“AI赋能科学”领域的初创企业都是先搭建模型,再搜集数据。英格瓦松在接受《财富》采访时称,Holyvolt走的是相反路径:“我们深耕强化学习20年,多年积累的研究成果,正是AI开展学习、迭代演进时可以依托的根基。”

“我们选择的是另一条技术路线,搭建这套体系难度极高,”他说,“需要经年累月的沉淀。”

英格瓦松表示,一项与客户共同完成的案例研究显示,该实验室的研发速度已达传统研发模式的10倍之多,而Holyvolt认为,一旦依托AI实现闭环迭代,相较于传统研发模式,这一速度还能继续提升。“我甚至可以说,我们已经找到了破解之道。”他说。

最后期限

该研发体系产出的首个成果是无序岩盐材料(disordered rocksalt,简称DRX),一类不含镍、不含钴的正极材料。普拉斯特称,其性能可媲美镍锰钴三元锂,“优于或等同于”镍基化学体系,在北美的落地成本与磷酸铁锂大致相当。他表示,无序岩盐的核心设计要求之一,是打造“完全安全可控”的供应链,但拒绝透露供应商信息。无序岩盐的性能数据已在2026年3月的国际电池大会(International Battery Conference)等多场学术会议上公布。

这一时间节点暗含一项硬性期限。2027年10月1日,美国《2024财年国防授权法》第154条将正式实施。该条款禁止美国国防部采购宁德时代、比亚迪以及另外四家列明的中国厂商生产的电池,而美军当下正着力扩大无人机的应用规模。普拉斯特称,Holyvolt计划在2027年实现无序岩盐材料的商业化落地,优先面向无人机领域,之后再拓展至交通出行市场。该公司同时在开发面向车企的电池配方,还在研发适配数据中心的专用方案(数据中心的GPU启动时,对瞬时峰值功率有极高要求)。

“这一幕我早见过”

对普拉斯特来说,这场较量有着特殊的个人意义。“我在密歇根北部长大,家族世代深耕汽车行业,”他说,“这一幕我早见过,结局如何,我心里有数。”

在他看来,这场较量更多关乎稳定,而非对抗竞争。“无论哪个领域,若资源过度集中于一处,局面就会失衡。” 他说。

英格瓦松也持类似观点。他说:“我对中国没有任何意见” ,并随即提到,他祖国最知名的车企沃尔沃,“如今已是中方旗下企业。”(中资车企吉利持有该公司多数股权。)

Holyvolt无意自建超级工厂。普拉斯特说,公司愿意开展技术授权、委托代工或合资合作,把工业化量产这部分业务留给合作伙伴。

从当前阶段走向真正工业化量产的这一步,才是最大难关。(财富中文网)

译者:梁宇

审校:夏林

Steven Prast is old enough to remember when a mobile phone came in a bag. It was too heavy to carry around, so most people mounted it in the car. Prast argued in a Zoom interview with Fortune that what eventually shrank it into something you wear on your wrist was more the battery than the transistor, holding up his own phone as proof.

“What took us from that to this is the battery,” said Prast, a former Tesla manufacturing leader who is now chief operating officer of the battery-materials startup Holyvolt.

By his reckoning, lithium-ion technology is at best 20% of the way through its maturity curve, maybe 10% by conservative estimates. The West won’t win the rest of that curve, he said, by fighting with China over the part that’s already finished.

“We will not be able to out-China China on the battlefield that China has defined,” Prast said. “They’re the world champions at being China.”

Holyvolt is putting $45 million behind the alternative, all in equity. The Stockholm-based startup told Fortune that it has raised the round from Fifth Wall, Zone II, CC.VC and several angel investors from the defense and finance industries to hunt for battery chemistries China doesn’t control. The round values Holyvolt at $245 million, according to a person familiar with the company.

The gap Holyvolt is trying to jump is enormous. China makes about 85% of the world’s cathode active material, the part of a battery that largely sets how much energy it stores and what it costs. It makes about 80% of the world’s battery cells. Its lead comes from scaling the two chemistries that dominate today, NMC and lithium iron phosphate, or LFP.

“If we take LFP and we try to out-LFP China, we will lose,” Prast said. “We’ll lose that every day.”

The answer, he said in a statement, is to “invent better materials and processes and make them in the West.” He told Fortune he thinks there must be “another dozen cathode materials that humanity just hasn’t figured out yet.”

A chemistry problem

If batteries are so far from maturity, why haven’t they followed the steady, predictable curve of computer chips? Prast’s answer is that Moore’s Law comes out of physics, a science humans understand well enough to solve on paper, and batteries belong to another tradition.

“If you design an airplane and you do that computationally, I will fly on that airplane,” he said. “With chemistry, there are still huge mysteries.”

“The problem for battery is that it’s a chemistry problem and not a physics problem,” he said. “If it were a physics problem, you’d have Moore’s Law.”

That puts Holyvolt on one side of a split in AI-driven science. Most of the field, Prast said, believes it can solve chemistry computationally. A smaller camp believes AI still needs a physical lab to check its work. Holyvolt is in the second camp.

“We have to go to the lab and we have to go test,” Prast said.

The bet

Holyvolt’s lab is Wildcat Discovery Technologies, a San Diego company it merged with in a deal The Next Web reported at $73 million in March, a figure the firm confirmed as accurate. Founded by Scripps chemist Peter Schultz, Wildcat built a high-throughput system that tests thousands of battery materials in parallel. Over 18 years, the company says, the platform has run more than 225 projects for more than 85 customers, including BMW, and tested more than 500,000 cells.

That pile of data is the asset Mathias Ingvarsson, Holyvolt’s Swedish founder and CEO, built the company around. Ingvarsson was one of the core drivers behind Tempur-Pedic’s international expansion and its success before and after its IPO on the New York Stock Exchange, and ran its European business for six years after the listing. He told Fortune that, as one merged company, Holyvolt has trained AI models on Wildcat’s results and wired them back into the lab, so each round of real-world tests sharpens the next prediction. The data and the lab underpin the AI models being built to deliver the predictions the market is seeking, as well as to speed up R&D and develop next-gen breakthrough chemistries.

Most AI-for-science startups start with models and then look for data. Ingvarsson told Fortune that Holyvolt came at it from the other direction, with “the reinforcement learning we have done for 20 years, which AI needs to actually learn and build on.”

“We’re coming from the other side, which is very, very difficult to build,” he said. “Take years and years and years.”

Ingvarsson said the lab already works 10 times faster than conventional R&D, based on a case study completed with a customer, but Holyvolt thinks closing the loop with AI will push that figure upward compared to conventional R&D. “I’d go so far as to say that we have cracked the code on how to do it,” he said.

The deadline

The first thing out of that loop is DRX, short for disordered rocksalt, a family of cathode materials that use neither nickel nor cobalt. Prast described it as delivering “NMC-like” performance, “better than or equal to” nickel-based chemistry, at roughly the landed cost of LFP in North America. A core design requirement, he said, was that DRX’s supply chain be “entirely friendly.” He declined to identify suppliers. DRX has published performance data at scientific conferences, including the International Battery Conference in March 2026.

The timing includes a built-in deadline. Starting Oct. 1, 2027, Section 154 of the fiscal 2024 defense authorization act bars the Pentagon from buying batteries from CATL, BYD and four other named Chinese manufacturers, even as the military pushes to expand its use of drones. Holyvolt plans to begin commercializing DRX in 2027, starting with unmanned aerial vehicles before moving into mobility. It is also working on formulations for automakers and one aimed at data centers, which need bursts of very high power as GPUs spin up, Prast said.

‘I’ve seen this movie before’

For Prast, the fight is personal. “I grew up in northern Michigan. Long family history in the auto industry,” he said. “I’ve seen this movie before. I know how it ends.”

He framed it as a matter of stability more than rivalry. “It’s unstable when you have too much concentration in any one place,” he said.

Ingvarsson took a similar line. “I have nothing against China,” he said. Then he noted that Volvo Cars, his home country’s best-known carmaker, “happen to be Chinese nowadays.” (China’s Geely owns a majority of the company.)

Holyvolt doesn’t plan to build gigafactories. Prast said the company is open to licensing, toll manufacturing or joint ventures, and would leave industrial scale to partners.

“The transition from there to actual industrial scale is the killer.”

财富中文网所刊载内容之知识产权为财富媒体知识产权有限公司及/或相关权利人专属所有或持有。未经许可,禁止进行转载、摘编、复制及建立镜像等任何使用。
0条Plus
精彩评论
评论

撰写或查看更多评论

请打开财富Plus APP

前往打开