英伟达预计销售额达 6730 亿美元，AI 需求持续扩大
英伟达预测 2028 财年增长 70%，这意味着销售额将达到 6730 亿美元——尽管供应受限，但需求已扩展到超大规模云厂商之外。
图片来源：Blockonomi
英伟达预计 2028 财年营收将增长 70%，如果当前华尔街对 2027 财年的共识预期成立，那么年度销售额将达到约 6730 亿美元。这将使这家芯片制造商在营收上超越苹果和 Alphabet，在美国科技公司中，预计销售额高于它的只剩亚马逊。
首席财务官 Colette Kress 于 2026 年 8 月 26 日发布了这一预测。这一数字远高于 LSEG 追踪的分析师平均预期（44%），也标志着英伟达信息披露方式的变化：该公司此前从未提供过如此长远的预测，尽管 CEO 黄仁勋曾给出过较短期内的 AI 芯片需求预期。
英伟达 2027 财年第二季度业绩为这一展望提供了直接佐证。季度营收达到 962 亿美元，是去年同期的两倍多，其中数据中心营收增长 117%，达到 890 亿美元。英伟达股价在盘后交易中一度上涨约 4%（据某家报道），而另一份市场报告称盘中高点为 5.6%。这一差异反映的是财报发布后不同的交易区间报道，并非公司预测发生变化。
眼下制约近期上限的是供应而非需求。黄仁勋表示，包括内存在内的零部件短缺，使英伟达无法做出更高的预测——因为 AI 基础设施正在消耗全球芯片和内存产能中越来越大的份额。
推荐阅读
Shein 在估值见顶后目标 IPO 募资 270 亿美元
Marcus Vance • 2026 年 8 月 25 日 • 阅读时长 3 分钟
“我们的需求远不止 70%。我们的供应能力让我们有信心交付 70% 的增长，我们将继续与供应链合作，在这个基础上进一步提升。”
英伟达正在扩大其客户基础
投资者一直担心，英伟达的增长过度依赖少数几家超大规模云服务商，而这些云服务商的数据中心又是为少数前沿 AI 实验室而建。黄仁勋表示，下一阶段将迎来更广泛的买家群体，包括区域性 AI 公司、新型云服务商（neocloud）、初创企业和传统企业。
英伟达将这些客户统称为 ACIE。该公司向它们销售的不只是 GPU。黄仁勋表示，其技术可以提供数据中心周边堆栈的很大一部分，使英伟达成为那些无法自行搭建整套系统的组织的更广泛基础设施合作伙伴。他说，这一客户类别此前“几乎隐形”，但现在包括越来越多将 AI 用于实际生产工作、而不仅仅是资助大规模模型训练的公司。
“去年这个时候，只有一家实验室在推动建设；今天，我们迎来了新 AI 实验室和初创企业的黄金时代，多个前沿实验室并行扩张，开源模型生态蓬勃发展，物理 AI 正在落地——在美国和全球都展现出强劲势头。”
客户基础分散在区域性服务商、初创企业和传统企业之间，可以降低对任何单一超大规模云服务商资本预算的依赖。这也让英伟达有更多机会在销售处理器的同时，配套销售网络、系统和其他数据中心组件，不过财报并未单独披露这些产品贡献了多少收入。
这种需求已经开始影响美国以外的公司。英伟达财报发布后，欧洲半导体股上涨：ASML 上涨约 2.5%，而 STMicroelectronics、Infineon Technologies 和 BE Semiconductor 各自上涨 2% 至 4%。这些走势表明，投资者预期用于建设 AI 基础设施的设备与组件支出将持续增长，尽管欧洲主要股指整体下跌或持平。
英伟达正在为购买其芯片的基础设施提供融资
英伟达的角色如今已不止于销售硬件。该公司正在投资包括 OpenAI 和 Anthropic 在内的模型开发商，支持租用英伟达算力的新型云服务商（neocloud），并帮助安排数据中心建设的融资。
一个例子是，为俄亥俄州正在建设的大型算力园区提供 1050 亿美元金融支持，OpenAI 预计将成为该园区的租户。Nvidia 还宣布与华尔街主要公司合作，安排高达 5000 亿美元的数据中心融资。这些安排引发了人们对循环融资的担忧：Nvidia 帮助客户或基础设施融资，而这些资金又可能通过购买其产品回流到 Nvidia。
黄仁勋为这一战略辩护，认为前沿 AI 公司在其资产负债表足以支撑低成本借贷之前，需要异常庞大的资本。
“这是第一代需要数百亿美元才能获得融资的初创公司。上一次听说有初创公司需要数十亿美元才能起步、需要数百亿美元才能盈利是什么时候？这种事从未发生过。但这确实是 AI 的本质。构建 AI 的成本、部署 AI 的成本，都是高度资本密集型的。”
他表示，这些公司中有许多不具备投资级评级，也缺乏获得低成本资本所需的运营和财务历史。Nvidia 希望投资它们、支持它们，并鼓励它们在 Nvidia 的技术上进行构建，从而将公司的财务风险敞口与其计算平台的未来需求绑定在一起。
“它们不具备投资级评级，没有足够的业绩记录——资本方面的记录、财务方面的记录——来以低成本获取或确保资本。而这正是 Nvidia 可以发挥作用的地方。”
黄仁勋还表示，如果某家 AI 公司出现问题，Nvidia 的基础设施可以在客户和工作负载之间重新部署，从而限制与单一借款人或租户绑定的风险。
“我们投入的资金将产生巨大回报。我认为风险很低。”
这是英伟达的辩护之词，并非对融资风险的独立评估。公司财报显示需求是真实的，但融资结构使英伟达既是供应商，也日益成为购买其系统所需资金的直接参与者。悬而未决的问题是，更广泛的 ACIE 客户群能否在英伟达不继续以自身资产负债表补贴扩张的情况下，维持预期的增长。
马库斯·万斯
企业版编辑
马库斯追踪资金流向。他报道企业软件、云架构以及大型科技公司战略中的结构性转变。他将晦涩的财报电话会议和复杂的并购活动，转化为关于行业实际走向的可操作洞察。如果某家科技巨头悄然转向，马库斯通常是第一个察觉的人。
/ 继续阅读
Nvidia projects $673 billion in sales as AI demand widens
Nvidia forecasts 70% fiscal 2028 growth, implying $673 billion in sales as demand expands beyond hyperscalers despite supply constraints.
Image: Blockonomi
Nvidia is projecting 70% revenue growth in fiscal 2028, which would put annual sales at roughly $673 billion if the current Wall Street consensus for fiscal 2027 holds. That would move the chipmaker ahead of Apple and Alphabet by revenue, leaving only Amazon among US tech companies with higher projected sales.
CFO Colette Kress delivered the forecast on August 26, 2026. It is far above the 44% average analyst estimate tracked by LSEG and marks a change in Nvidia’s disclosure: the company has not previously provided a forecast this far into the future, although CEO Jensen Huang has offered shorter-range indications of expected AI chip demand.
Nvidia’s fiscal 2027 second-quarter results provided the immediate evidence for the outlook. Quarterly revenue reached $96.2 billion, more than double the year-earlier figure, while data-center revenue rose 117% to $89 billion. Nvidia’s shares climbed about 4% in extended trading in one account, while another market report put the session’s high at 5.6%. The difference reflects the trading range reported after the earnings release, not a change to the company’s forecast.
Supply, rather than demand, is now setting the near-term ceiling. Huang said shortages in components including memory prevented Nvidia from making a still higher projection as AI infrastructure consumes a growing share of global chip and memory capacity.
Recommended reading
Shein targets a $27 billion IPO after its peak valuation
Marcus Vance • Aug 25, 2026 • 3 min read
“Our demand is much greater than 70%. Our supply allows us to confidently deliver 70%, and we’re going to continue to work with our supply chain to increase on on that.”
Nvidia is broadening its customer base
Investors have worried that Nvidia’s growth depends too heavily on a small group of hyperscalers building data centers for a handful of frontier AI labs. Huang said the next phase will involve a wider set of buyers, including regional AI companies, neocloud providers, startups and conventional enterprises.
Nvidia groups those customers under the label ACIE. The company is selling them more than GPUs. Huang said its technology can supply much of the surrounding data-center stack, making Nvidia a broader infrastructure partner for organizations that cannot assemble the system themselves. The customer category was previously “largely invisible,” he said, but it now includes a growing number of companies deploying AI for useful production work rather than merely funding large-scale model training.
“This time last year, one lab alone was driving the buildout; today, we have a golden age of new AI labs and startups, multiple frontier labs scaling in parallel, a thriving open-model ecosystem and physical AI coming online — with strong momentum across the U.S. and around the world.”
A customer base spread across regional providers, startups and enterprises could reduce dependence on any one hyperscaler’s capital budget. It also gives Nvidia more opportunities to sell networking, systems and other data-center components alongside its processors, although the supplied results do not break out how much revenue comes from those products.
The demand is already affecting companies outside the US. European semiconductor stocks rose after Nvidia’s results: ASML gained about 2.5%, while STMicroelectronics, Infineon Technologies and BE Semiconductor each rose between 2% and 4%. The moves suggest investors expect continued spending on the equipment and components needed to build AI infrastructure, even as broader European indexes declined or held flat.
Nvidia is financing the infrastructure that buys its chips
Nvidia’s role now extends beyond selling hardware. The company is investing in model developers including OpenAI and Anthropic, backing neocloud providers that rent Nvidia-powered compute, and helping arrange financing for data-center construction.
One example is $105 billion in financial support for a large compute campus under construction in Ohio, where OpenAI is expected to be the tenant. Nvidia has also announced a partnership with major Wall Street firms to arrange up to $500 billion in data-center financing. Those arrangements have raised concerns about circular financing: Nvidia helps fund customers or infrastructure, and that money can then flow back to Nvidia through purchases of its products.
Huang defended the strategy by arguing that frontier AI companies need unusually large amounts of capital before they have the balance sheets required to borrow cheaply.
“This is the first generation of startups that needed tens of billions of dollars to get funded. When was the last time anybody heard of a startup that needed billions of dollars to get off the ground and needed tens of billions of dollars to become profitable? That just never happened. But that’s really the nature of AI. The cost of building AI, the cost of deploying AI, it’s very capital intensive.”
He said many of those companies are not investment grade and lack the operating and financial history needed to secure low-cost capital. Nvidia wants to invest in them, support them and encourage them to build on Nvidia’s technology, tying the company’s financial exposure to future demand for its computing platform.
“They’re not investment grade. They don’t have the track record, the capital track record, the financial track record, to be able to capture or secure capital at a low cost. And this is where Nvidia could be helpful.”
Huang also said Nvidia’s infrastructure can be redeployed across customers and workloads if an individual AI company falters, limiting the risk of being tied to one borrower or tenant.
“The money we’ve invested is going to generate tremendous returns. I think the risk is low.”
That is Nvidia’s defense, not an independent assessment of the financing risk. The company’s earnings show that demand is real, but the financing structure makes Nvidia both a supplier and an increasingly direct participant in the capital required to buy its systems. The unresolved question is whether the broader ACIE customer base can sustain the projected growth without Nvidia’s balance sheet continuing to subsidize the expansion.
Marcus Vance
Enterprise Editor
Marcus follows the money. He covers enterprise software, cloud architecture, and the tectonic shifts in Big Tech strategy. He translates dense earnings calls and complex M&A activity into actionable insights about where the industry is actually heading. If a tech giant makes a silent pivot, Marcus is usually the first to notice.