DeepSeek Leads AI Asset Revaluation in China

Advertisements

The allure of China’s artificial intelligence (AI) assets is becoming increasingly apparent, with leading AI companies positioned to seize performance-driven alpha investment opportunitiesValuations of these top-tier companies still exhibit a significant comparative advantage relative to their American counterpartsAs pressures from a slowdown in capital expenditures (CAPEX) impact both performance and valuations in the US tech sector, the attractiveness of Chinese assets is anticipated to further intensifyAnalysts maintain a long-term optimistic outlook for the revaluation and growth potential of China's technology assets in a broader contextEach breakthrough in technology and its application presents an opportunity for systemic valuation enhancementInvestment opportunities should be concentrated in homegrown large models, intelligent cloud solutions, AI data centers (AIDC), domestic computing processes, autonomous driving, AI agents, and embodied intelligence.

A wave of enthusiasm surrounds the emergence of DeepSeek, as investors ponder the breadth of revaluation for Chinese tech assetsExhibiting remarkable performance, significantly reduced costs, and an open-source ecosystem, DeepSeek has set off a 'butterfly effect' within the industry.

In a mere seven days, DeepSeek amassed over 100 million registered users, shattering the record previously held by ChatGPT, which took two months to achieve the same milestoneAs of now, Daily Active Users (DAU) for DeepSeek have surpassed 30 million, making it the most popular AI large model globally and catalyzing a comprehensive reassessment of Chinese AI assets.

A retrospective examination of the impact following the launch of ChatGPT reveals a substantial revaluation among American tech firmsSince the beginning of 2023, the Nasdaq index has soared by 100%. Notably, Nvidia's stock price skyrocketed by 8.5 times, propelling its market capitalization into the global top threeMeanwhile, application companies such as Palantir and Applovin also experienced phenomenal surges, climbing by 17.6 times and 47.4 times respectively

Advertisements

Major players in China like Alibaba, SMIC, BYD, and Xiaomi saw their stock prices increase by 51%, 43%, 37%, and 30% since JanuaryIDC and cloud infrastructure companies also exhibited considerable price increases as the valuation gap amongst tech firms began to narrowDeepSeek plays a pivotal role in transforming the narrative of American AI dominance, thereby shifting global investor sentiment regarding the value of Chinese tech assetsThe breakthrough from this startup, founded just in July 2023 and currently employing slightly over 100 individuals, encapsulates the essence of China’s technological innovation.

The industrial implications of DeepSeek are profound: it heralds an era of AI egalitarianism, reshaping the global technological landscape, increasing demands for computing power, and igniting expectations for an application explosion.

Operating under constrained computing power, DeepSeek has pioneered the replication of GPT-01, achieving performance benchmarks in mathematics, coding, and natural language that rival GPT-01. The independent open-source RL-Scaling technology facilitated the extraordinary feats of this modelThe V3 version incorporates innovative architectures such as Mixture of Experts (MoE) and Multi-Layer Architecture (MLA), along with revamped training strategies, enabling a cost-effective and efficient training regimenAccording to DeepSeek-R1 documentation, the pre-training costs amounted to less than $6 millionFurthermore, the R1 version explores transitioning from supervised fine-tuning (SFT) to zero-supervision fine-tuning, employing a combination of Generalized Reward Propagation Optimization (GRPO) and an efficient reward system, yielding performance improvements while drastically reducing costsDeepSeek effectively dismantles the monopolistic advantage held by American firms in the AI race; its promotion of AI solidarity is poised to alter the global tech industry dynamic, further stoking the anticipation of application breakthroughs.

Currently, AI applications find themselves on the precipice of an explosive growth phase, with computing demands poised for revival

Advertisements

Companies such as ByteDance’s Doubao and Alibaba’s Tongyi Qianwen are expected to rapidly follow suit in innovation, leading to a flourishing variety of vertical models.

The race among modeling companies drives a ‘pulse-like’ growth in demands for training computing power, while needs for inference computing power and local deployments of large corporate models are also set to experience exponential growthObservers maintain a positive outlook on investment opportunities in AI cloud computing, AIDC data centers, AI chips, and advanced manufacturing processesOn the application front, scenarios such as autonomous driving, AI advertising, AI agents, AI smartphones, and embodied intelligence are progressively expected to materialize.

Looking further ahead, the future of AI does not solely hinge on creating more powerful models; rather, it resides in solving real-world problems through AI applications and generating economic value in the process.

From technological innovation to real-world application, from shifting expectations to generating tangible value, the journey for Chinese tech enterprises remains lengthyAt present, American model companies continue to significantly lead the way, retaining robust advantages in cutting-edge fields such as multi-modal models, world models, and reinforcement learningLooking forward, the gap in AI technology between the US and China is likely to persist, though China stands to uncover vast opportunities by synchronizing AI technology with traditional industrial foundations to address practical issues and generate economic value.

It is posited that applications in intelligent driving, AI advertising, and AI agents will likely be among the first to take root, contributing revenue increments and altering market dynamics starting from 2025. Apple's iPhone 17, launching in the third quarter, showcases innovations that could bring about a new wave of on-device intelligence opportunities, with local models from Alibaba’s Tongyi Qianwen likely to play a role therein

Advertisements

Embodied intelligence aligns well with China’s solid manufacturing base and may emerge as another competitive sector for the nationFrom an industry application standpoint, the intersections of AI with sectors such as internet, finance, healthcare, industry, and transportation are anticipated to materialize firstAnalyzing the business value chain, the intersections of AI with auditing, regulations, and research may significantly enhance operational management efficienciesIn the more distant future, an intricate interconnection and mutual influence between the physical and digital domains beckons, potentially accelerating the advent of the metaverse.

Regarding the revaluation logic, the trajectory of R&D expenditures within the tech sphere is shifting from “negative valuation” to “positive valuation.”

In the post-pandemic era, China’s tech assets remain persistently undervalued, primarily due to two factors: firstly, a slowdown in earnings growth has bred uncertainty among investors regarding core profit projections; secondly, China’s AI sector is often perceived as “generationally lagging,” leading to the characterization of new business R&D investments as “losses” that encroach upon valuationsIn recent years, Chinese tech giants have frequently been appraised based on aggregate price-to-earnings (PE) valuations, reflecting the losses induced by new business R&D in a negative valuation frameworkThe reevaluation prompted by DeepSeek superficially appears to elevate the PE multiples for leading firms, yet fundamentally signifies a transformative shift for Chinese AI assets from negative to positive valuationIn the future, should economic expectations improve, there exists scope for enhancements in core business valuation multiplesShould the AI segment register substantial revenue contributions, realignment toward price-to-sales (PS) valuation across segments is also plausibleAnalysts assert that the revaluation of Chinese technology assets is still in its infancy

Advertisements

Advertisements

Post Comment