Nvidia, XPeng, Alibaba: The $30 Billion Bet on the AI Application Layer
Three titans shift focus from hardware to the next layer of AI dominance
Three distinct developments in artificial intelligence landed on the same Monday from three different corners of the global economic map. Each move points in the same direction. The race for AI dominance has moved decisively beyond the question of who manufactures the best chip. The current contest is now over who controls the applications robots and the multimodal models that will run on those chips over the next decade. This shift represents a fundamental change in how investors view technology value.
Nvidia the California chipmaker whose graphics processing units have provided the physical infrastructure for the current AI boom is in discussions to invest in AI search startup Perplexity as part of an equity financing round that would value the company at more than $30 billion. The Information reported Sunday citing people familiar with the matter. The funding round is expected to amount to several billion dollars. This represents a growth of more than 50 percent from Perplexity's valuation a year ago. The discussions were not confirmed by either party as of Monday morning. This potential investment follows a period in which Nvidia reportedly explored paying billions to license Perplexity's technology and acquire specific technical talent before pivoting toward a direct equity stake. This sequence illustrates how seriously Nvidia's leadership is treating the transition from infrastructure provider to participant in the AI application layer.
The Perplexity discussions are not Nvidia's first or largest application-layer bet of recent weeks. Earlier this month the company committed $6 billion to a deal involving AI startup Poolside. The goal was to build one of the world's most powerful open-weight AI models. Open-weight models refer to systems whose parameters are publicly available. These systems can be downloaded customized and run by any developer company or government without requiring access to proprietary APIs. Jensen Huang has become one of the most vocal advocates for open-weight AI in the technology industry. He organized a letter to Washington urging against restrictions on open models. This letter quickly gathered dozens of corporate signatories including OpenAI Alphabet Microsoft Meta Dell Palantir and IBM. The strategic logic behind Huang's open-weight advocacy is not merely principled. By IDC's May 2026 estimate Nvidia controls approximately 81 percent of the data-center AI chip market. Closed AI models concentrate GPU demand inside a handful of well-funded labs. Open-weight models distribute that demand across the entire economy. Startups enterprises governments and research institutions each require Nvidia compute to run whatever models they build or deploy. "Whenever there will be more use you'll have to sell more Nvidia computers we'll have to build more data centers and the technology will diffuse into more industries" Huang told Axios. OpenAI is from Nvidia's perspective a market expansion strategy wrapped in a principled framing. Nvidia forecasts $1 trillion in total sales for its Blackwell and Vera Rubin chip architectures across 2026 and 2027. That number cannot be reached without the broadest possible base of AI adoption globally.
Perplexity has built an AI-native search engine designed to answer questions conversationally rather than return ranked links. The company has seen its annualized revenue surge from less than $250 million in early 2026 to more than $750 million. This growth is driven significantly by the adoption of Perplexity Computer. Perplexity Computer is an AI agent product that can take actions on a user's behalf rather than simply answering questions. For Nvidia an investment in Perplexity would represent a deliberate move to secure a stake in the inference layer. The inference layer is the point at which AI models interact with users and generate ongoing revenue. The boundary between the company that sells AI infrastructure and the companies that build AI applications is blurring. Nvidia's combined bets on Poolside and Perplexity represent a coherent strategy to occupy both sides of that line simultaneously. The strategy involves building open-weight models that expand the ecosystem while investing in the consumer-facing applications that monetize it.
Simultaneously on Monday Chinese electric vehicle maker XPeng announced it had raised more than $900 million in new capital specifically dedicated to humanoid robot development. This is the largest single fundraising dedicated to that category by a Chinese automotive company and one of the largest globally. XPeng's humanoid robot program is centered on its Iron robot model. This robot model stands 1.72 meters tall and uses 60 joints capable of 200-degree motion ranges. The capital injection comes as the company positions itself alongside Tesla's Optimus program Figure AI and other Western humanoid competitors. XPeng's chairman has said the company plans to begin mass production of its Iron robot and flying car models in 2026. These targets are the ones that the $900 million fundraising is designed to support. The fundraising arrived the same day that two Chinese humanoid robots surpassed Usain Bolt's long-standing men's 100-meter world record of 9.58 seconds during the opening of the second World Humanoid Robot Games in Beijing. This was a symbolic milestone that Chinese state media covered extensively as evidence of the country's accelerating capabilities in physical AI.
XPeng's parent company has Alibaba as a significant shareholder. The e-commerce and cloud giant holds approximately 7.5 percent of XPeng. This cross-shareholding structure connects two of Monday's three major AI announcements and reflects the degree to which China's technology ecosystem is more deliberately integrated across companies than its American counterpart. Alibaba completed the trio with the formal commercial launch of Wan3.0. Wan3.0 is the newest version of its generative AI video model. The company announced this release one day after announcing a HKD 80 billion share placement to fund AI infrastructure expansion. Wan3.0 represents a significant capability expansion from its predecessor. Earlier versions of the model required users to input text or image prompts to generate video. Wan3.0 can accept documents spreadsheets presentations PDFs and web content as inputs. It converts existing business materials directly into finished video media. Public beta access runs through Alibaba Cloud's Model Studio and Qwen Cloud platforms. The commercial logic behind the breadth of accepted input formats reflects a broader shift in how AI platforms are positioning themselves for enterprise adoption. Rather than asking corporate users to learn new prompting techniques Wan3.0 treats the documents organizations already have as the raw material for AI-generated content.
Alibaba's video model launch arrives in a context of intensifying competition between Chinese and American AI labs. Chinese technology companies including Baidu ByteDance Tencent and Alibaba have spent the past 18 months releasing increasingly capable open-weight models. These models benchmark close to the performance of the best Western closed-source products. The strategy of releasing open-weight models serves multiple purposes simultaneously. It builds developer communities establishes technical credibility and places pressure on American companies by making frontier-adjacent AI capability available at no cost. Alibaba's Wan 2.1 family released in early 2025 demonstrated that the company could produce video generation capability comparable to Western competitors and distribute it freely. This pattern Wan3.0 extends with substantially more powerful multimodal input handling. Chinese labs have spent two years driving AI inference costs down. The floor keeps falling as domestic rivals ship trillion-parameter systems that benchmark close to their Western rivals. The convergence of these three developments reveals a clear picture. The industry has moved past the hardware bottleneck. The next layer of value lies in the software the robots and the models that define how artificial intelligence integrates into daily life. This transition requires massive capital allocation and strategic coordination among the world's largest technology companies.
First reported by hngn.com.
More in Business
- IMF Chief Warns of Economic Divergence and Debt Risks Amid Artificial Intelligence Boom
Kristalina Georgieva suggests that while AI could significantly boost global growth, it simultaneously threatens to widen inequality and exacerbate inflationary pressures and public debt burdens.
- California Limits Utility Profits and Expands Home Energy Role
Gov. Gavin Newsom signed laws capping utility returns on lower-risk projects and accelerating virtual power plant development.
- Lyft to Pay $272.5 Million in Record California Wage-Theft Settlement
Largest such settlement in state history resolves claims that Lyft misclassified drivers as independent contractors from 2016 to 2020, denying them core wage protections.