Qiming Venture Partners' Alex Zhou: At the 'Everest Base Camp,' Foreseeing the Path to Kingship for AI Startups
Artificial intelligence is undergoing a profound shift from "bolt-on" to "built-in," from spectacle to commercial deployment.
Editor's Note: The AI industry is entering a new cycle — moving from technical spectacle to commercial兑现. Investment in computing infrastructure continues to heat up, with breakthroughs in models, embodied intelligence, and agent technology. Yet the application layer shows a polarized landing pattern, and new business models adapted to the AI ecosystem remain to be formed. Coinciding with the 25th anniversary of the Forbes Midas List, Alex Zhou, Managing Partner of Qiming Venture Partners, was named to the 2026 Forbes Midas List of the World's Best Venture Capitalists. Forbes China conducted an in-depth interview with Zhou, in which he drew on the firm's years of full-chain AI investment布局 and its "half-step ahead" investment methodology to deconstruct the core competitive logic of the AI infrastructure construction period, predict key inflection points in the agent industry, and deeply analyze core topics such as the valuation reconstruction of primary and secondary markets, listing choices for hard-tech enterprises, and the cyclical patterns of the technology industry. This article is a transcript of the Forbes China interview, reprinted with authorization on the Qiming Venture Partners WeChat official account.
On July 19, at the World Artificial Intelligence Conference, Qiming Venture Partners released its "Top 10 AI Outlooks for 2026," sketching a picture of artificial intelligence undergoing a profound transformation from "external add-on" to "internal engine," from performance to commercial landing. At the technical level, large models are internalizing capabilities that were once "external add-ons" — task planning, tool invocation, and multi-agent collaboration — while multimodal models are evolving toward interactive "world models," enabling AI to truly learn to perceive and plan the physical world. Effective data at leading robotics companies will leap from the "ten-thousand-hour level" to the "million-hour level"; dexterous hands with tactile sensing will mature and continue to decline in cost, forming a high-low pairing with two-finger gripper solutions and gradually achieving scaled penetration in complex manipulation scenarios. The infrastructure story more closely resembles an arms race. The center of gravity of AI computing demand is shifting from training to inference; storage, advanced process, and packaging capacity are tightening at every layer; computing reserves are being upgraded to core strategic assets for AI enterprises; and competition has escalated from single-point chips to system-level competition encompassing interconnect, thermal management, and power delivery. Safety and trustworthiness have been upgraded from an option to a necessity, becoming one of the three key variables for large-scale enterprise AI deployment alongside product efficacy and token cost. Most intriguing is the rewriting of commercial logic. The freemium model of the Internet era is accelerating its departure, replaced by outcome-based and value-based pricing — the measure of an AI company is no longer user scale, but how much commercial value each unit of intelligence cost can create. Efficiency tools will explode before entertainment applications, and true "AI-native organizations" will move from concept to empirical proof: the per-capita output of a cohort of enterprises may be several times that of traditional organizations. To sum up this outlook in one sentence: AI is bidding farewell to its adolescence of spectacle, and industrial competition is henceforth shifting to a hard-power contest centered on financial operations and asset returns.
In early 2026, Alex Zhou, Managing Partner of Qiming Venture Partners, experienced an intensive harvest period in his investment career. Biren Technology, Zhipu AI, Axera, and Yunyinggu Technology — four projects he was deeply involved in — listed in succession. Zhou clearly understood that IPO is not the finish line, but merely arrival at Everest Base Camp at 5,200 meters. At several corporate IPO dinners, he used the Everest climb as a metaphor for this journey: "Reaching 5,200 meters is already remarkable, but the harder challenge is the subsequent push toward the 8,848-meter summit."
As a history enthusiast, Zhou maintains vigilance toward the historical depth of capital markets. He often cites NVIDIA as an example. When it went public in 1999, there were no shortage of rivals on par with it — 3dfx, ATI, and Matrox all stood at the same starting line. What allowed NVIDIA to pull away was not its listing-day pricing, but the technological leadership established by GeForce3 in 2001, and CUDA's "liberation" of parallel computing power from graphics to all domains in 2006. "It was precisely these post-IPO sustained technological innovations and major breakthroughs that opened NVIDIA's path to dominance," Zhou said. "I hope the AI companies we've invested in can also embark on their own paths to dominance after going public."
This year marks the 25th anniversary of the Forbes Midas List. Over these 25 years, the global venture capital industry has crossed two major waves — the Internet and mobile Internet — and now stands on the shore of the artificial intelligence wave. In Zhou's view, any large-scale technology wave exhibits a similar rhythm: the first phase is the infrastructure construction period, with focus on underlying technology breakthroughs and laying the foundational groundwork; once the foundation is solidified, a second phase arrives — longer in cycle and larger in scale — of comprehensive application deployment. Over the past 25 years, the vast majority of global VC energy was concentrated in the second phase of the Internet and mobile Internet, when applications exploded across all industries and enterprises' core competitiveness centered on extreme user and market insight capabilities and top-tier product-building capabilities. The new AI cycle that has just begun is clearly still in the first phase of the wave — the infrastructure construction period. "The core主线 of industry competition at this stage is the sustained breakthrough and iteration of underlying original technologies," he said. "This also places very different demands on investors' capabilities and cognition."
In Q1 2026, total financing in China's AI field approached three times that of the same period in 2025. Zhou believes: "When capital markets are in a downturn and rational stage, capital acts more like a filter — only a minority of outstanding enterprises with突出 strength and ample potential can obtain funding. Once markets fall into狂热, capital becomes a泡沫助推器." Therefore, after investing in over a hundred AI enterprises, Qiming Venture Partners has turned its attention to "less crowded corners" — scientific foundation models, agent frameworks and Harness Engineering, AI safety, vertical-domain agents, and AI-driven next-generation consumer hardware. "We want to unearth early-stage startups like Biren Technology in 2019 and Zhipu AI in 2021."
01 / AI Application Landing: Polarized
Zhou believes that over the past year, the landing of AI applications at the application layer has shown a significant polarized pattern.
Areas where landing has significantly exceeded expectations are concentrated in vertical applications, efficiency tools, and改良式 innovative consumer scenarios. Their core is: leveraging AI to optimize existing business formats,盘活存量 data, making value quantifiable and input-output logic clear. For example, intelligent agents in financial investment research have continued last year's strength; AI coding tools, intelligent documents, and meeting assistants have rapidly moved from pilot to scaled adoption. More representatively, AI+ dating and social platforms,借助 AI intelligent matching capabilities, have made previously high-cost manual matchmaker services accessible, quickly closing commercial loops. In content production, AI short drama production has become a phenomenon-level case, with production efficiency for equivalent-quality content improving by roughly 10x.
Areas where landing has been slower than expected are the Kill Time entertainment-oriented consumer applications that had the largest scale in the mobile Internet era. In AI entertainment products, no全民爆款 application has yet emerged; most remain at the concept or niche pilot stage, with乏力 growth.
Zhou identified three practical obstacles: First, the cost structure has fundamentally changed. Mobile Internet's marginal cost was nearly zero, while every AI call consumes computing power and tokens,叠加 customer acquisition costs — making it difficult for most C-end applications to cover expenses with revenue. Second,巨头 barriers are too high; giants hold computing power, data, and traffic, while startup teams struggle to differentiate, leading to severe homogeneous内卷. Third, business models lag behind technological development; the industry is still using traditional monetization methods, and no new commercial logic adapted to the AI ecosystem has yet emerged. "The explosion of next-generation AI-driven C-end applications requires a massive transformation in business models to配合," he emphasized.
Half-Step Ahead: Fast but Not Early, Investing in What You Understand
From perceptual intelligence to generative AI, and on to embodied intelligence, Zhou has consecutively hit several key technological nodes. He attributes this to Qiming Venture Partners' "half-step ahead" strategy.
"This precise, forward-looking rhythm does not rely on intuitive博弈, but comes from a mature methodology沉淀 through our years of实战," Zhou explained. The core of this methodology is: precisely锚定 technological development nodes, completing布局 after the technology breakthrough point but before the market explosion point. "We only enter after technology has completed iterative convergence, development goals are clear, and the industry has formed a mainstream technical implementation路线, avoiding the risk of investing too early during periods of technological路线 uncertainty; at the same time, we insist on positioning before market consensus forms and capital overheats, avoiding赛道 valuation泡沫."
What supports this approach is the "capability circle" and "friendship circle." The capability circle means sustained, in-depth industry research, being honest with oneself, and building solid professional cognition; the friendship circle is依托 top-tier industry resources to build a立体化 cross-verification system, avoiding single-source information bias and precisely identifying the true rhythm of industries. He believes investment institutions must acknowledge their capability boundaries: "This method only applies to fields we have deeply cultivated and thoroughly understood, where we possess both circles. Only invest in what you truly understand, don't blindly follow trends — understand means understand, don't understand means walk away."
02 / The Next Inflection Point: Agent-Driven
In Zhou's view, the AI industry has already crossed its first key inflection point. Marked by ChatGPT's release at the end of 2022, through the载体 of conversational robots, the marginal cost of knowledge acquisition was greatly reduced — though still far from approaching zero. For example, it can help you summarize a paper, plan a trip,串联,梳理,提炼, and interpret information into knowledge.
"Looking ahead one to two years, the next inflection point will be driven by AI Agent," Zhou judged. "People will be able to obtain deep-level cognition through Agents, thereby greatly reducing the marginal cost of task execution in the digital world, because Agents will do most of the work in the digital world. In the long term, as robotics and embodied intelligence technology develop, the marginal cost of executing various tasks in the physical world will also be gradually greatly reduced — the work in the physical world will also be done by robots. Humanity will迎来 comprehensive productivity革新, and social production paradigms will be thoroughly重构. Current Harness Engineering, world models, and multimodal understanding and generation integrated models are all technical explorations advancing toward these goals."
From Megvii in the AI 1.0 era, to Zhipu AI, Biren Technology, and Axera in the AI 2.0 era, Zhou has consecutively hit multiple key technological inflection points and invested in a batch of outstanding enterprises.
On the occasion of the 25th anniversary of the Forbes China Midas List, he engaged in a conversation with Forbes China about cycles, inflection points, and methodology. Below is an edited excerpt of the Q&A:
Forbes China: After DeepSeek's emergence, the industry has shifted significantly toward lightweight models and efficiency priority. In retrospect, what is its most long-term, most fundamental impact on China's large model industry? What will be the core of future model competition?
Alex Zhou: At this stage, pursuing higher model intelligence levels and breaking through technological capability ceilings remains the industry's core focus and will not be weakened. But as AI applications scale, industrial logic has undergone a key change: cost-performance has become a core metric equally important as intelligence capability. Once you have scale, cost issues become显性. The core competitiveness of future leading model labs will definitely be a "both/and" balancing capability: both持续攻坚 model architecture and algorithms to improve model capabilities and defend the technological ceiling; and through architecture optimization and engineering innovation, achieve lower training and inference costs. Future large model competition will no longer be a single-performance contest, but a comprehensive competition of intelligence capability, landing efficiency, and cost-performance — only then can the industrial closed loop truly be跑通.
Forbes China: Hong Kong's Chapter 18C and Tech Enterprise Dedicated Line have opened channels, but the market has seen a phenomenon of "more IPOs, lower valuations." For hard-tech enterprises, what are your core listing criteria? Would you advise "list fast if you can" or "solidify fundamentals and wait for the right timing"?
Alex Zhou: Hong Kong's 18C and Tech Enterprise Dedicated Line have indeed opened more convenient listing channels for hard-tech enterprises, but the market phenomenon of "more IPOs, lower valuations" is essentially value回归 under supply-demand imbalance. When large numbers of enterprises集中涌入, capital will more rationally筛选标的 that truly possess core moats; companies lacking fundamental support will sooner or later be voted down by the market. For hard-tech enterprises, my core thinking is: IPOs grow organically, they shouldn't be forced through赶鸭子上架.
Forbes China: Primary market valuations are returning to rationality. How do you believe the value anchor of future tech investment will be reconstructed? What kind of companies can maintain valuation resilience in a down cycle?
Alex Zhou: First, I want to express a core view: capital markets themselves have cycles, with alternating heat and cold, but this cycle actually has very little correlation with the intrinsic growth of an excellent tech enterprise. When markets rise, capital amplifies optimistic expectations and valuations across the board rise with the tide; when markets fall, capital turns conservative, valuations are broadly pressured and differentiation intensifies. But cycles are just short-term external environments — truly valuable enterprises will stand out regardless of market conditions.
Forbes China: The current VC "fundraising-investing-exiting" cycle is warming. Do you judge this as emerging from the valley bottom or the starting point of a new upward cycle? Set a theme word for China tech investment from 2026–2030 other than AI, and explain your reasoning.
Alex Zhou: The current venture capital market is comprehensively warming. In my view, the industry began emerging from the valley bottom from mid-2025 and is currently in a new upward phase. Looking ahead to 2026 through 2030, if selecting another core tech investment theme word besides AI, I would choose hard tech. Over the past two decades, the development and progress of numerous technology fields in China has核心依托 on continuously refined engineering capabilities. We possess a complete manufacturing system and mature supply chain,依托 a massive domestic market,叠加 a vast global market, continuously testing and solidifying our engineering capabilities.
Meanwhile, domestic basic scientific research strength has also improved over these two decades, with original innovation成果 continuously emerging. Based on such accumulation, over the next five years, the hard tech industry will form a new development paradigm, which I summarize as: scientific research strengthening, manufacturing amplification, and scale verification. We believe that in numerous细分 domains, we will create a batch of enterprises with global competitiveness.
Past Reviews
Qiming Weekly | Top 10 News from Qiming Venture Partners Portfolio Companies Vol.25, 2026.8.11
Qiming Venture Partners was founded in 2006. Currently, Qiming Venture Partners manages 11 USD funds and 7 RMB funds, with total assets under management reaching $9.5 billion. Since its establishment, it has focused on investing in outstanding enterprises in the early and growth stages of Technology and Healthcare innovation.
To date, Qiming Venture Partners has invested in over 580 high-growth innovative enterprises, of which more than 210 have listed on the New York Stock Exchange, NASDAQ, Hong Kong Exchanges and Clearing Limited, Shanghai Stock Exchange, and Shenzhen Stock Exchange, or exited through M&A and other means, with more than 80 enterprises becoming recognized unicorns or super-unicorns in their industries.
Many of Qiming Venture Partners' portfolio companies have grown into the most influential companies in their respective fields, including Xiaomi (01810.HK), Meituan (03690.HK), Bilibili (NASDAQ:BILI, 09626.HK), Zhihu (NYSE:ZH, 02390.HK), Roborock (688169.SH), Hesai Technology (NASDAQ:HSAI, 02525.HK), UBTECH (09880.HK), WeRide (NASDAQ:WRD, 00800.HK), HyperStrong (688411.SH), Insta360 (688775.SH), Unisound (09678.HK), Biren Technology (06082.HK), Zhipu AI (02513.HK), Gan & Lee Pharmaceuticals (603087.SH), Tigermed (300347.SZ, 03347.HK), Zai Lab (NASDAQ:ZLAB, 09688.HK), CanSino Biologics (688185.SH, 06185.HK), Schrödinger (NASDAQ:SDGR), MicroPort EP MedTech (688617.SH), Sanyou Medical (688085.SH), Amoy Diagnostics (300685.SZ), SinoCellTech (688520.SH), Insilico Medicine (03696.HK), Hope Medicine, Yuanxin Technology, MediLink Therapeutics, LaNova Medicines, StepFun, and others.