The investment strategy in the field of artificial intelligence is changing: the focus is shifting from celebrating increased efficiency to an arms race in computational power. The Kimi K3 model with 2.8 trillion parameters has become a turning point, comparable to the moment of DeepSeek 2.0.
After DeepSeek-R1 caused panic in the markets by suggesting that cheaper training would devalue investments in computing, the situation has changed. Jevons Paradox demonstrates that increased efficiency leads to an increase in overall consumption of computational resources, rather than a decrease in demand.
The volume of capital expenditures exceeds previous expectations. Google has increased its CapEx forecasts for 2026 by $15 billion, bringing them to the range of $195–$205 billion. Company management warns that infrastructure costs will continue to pressure margins due to rising depreciation and data center operating expenses, and they are expected to increase significantly in 2027.
Analysts predict that the total capital expenditures of the five largest global cloud service providers will exceed $800 billion in 2026, potentially reaching $1–$1.2 trillion by 2027. Meanwhile, the top 14 providers show an annual CapEx growth exceeding 90%, as the imbalance between supply and demand for AI computing power persists despite efficiency advancements.
The market has clearly divided between AI model developers and AI infrastructure providers. In June 2026, hyperscaler stocks fell by a cumulative 18%, marking the worst monthly performance since Meta's IPO in 2012, as investors questioned the possibility of proportional returns on massive capital investments. Even Google's earnings report, which exceeded expectations, triggered a 4% drop after the close of trading solely due to CapEx news.
Conversely, shares of AI hardware manufacturers rose by 16% during the same period, creating a mirror image of the trend. The most obvious winners were memory and semiconductor manufacturers, as the demand for HBM, DRAM, and GPUs remains structurally undersupplied.
The K3 model with 2.8 trillion parameters requires at least 10 GB300 GPUs for local deployment. Even if training costs decrease by 40%, doubling the training frequency within the arms race will lead to a 20% increase in net compute demand. Thus, efficiency stimulates more competition, not less.
The most acute bottleneck is memory: the supply of HBM from SK Hynix is limited for the next two to three years. AI inference is driving a resurgence in central processors, projected to grow 4–5 times. A Chinese example demonstrates the Zhipu AI data center with a capacity of 1 GW, showing that even companies focused on model efficiency are now actively investing in scaling computational power.
The release of Moonshot AI's Kimi K3 model has triggered a global reassessment of stocks in the artificial intelligence sector. Analysts call this the second shock following DeepSeek, forcing a revision of assumptions about the AI stock market, as the 2.8 trillion parameter model demonstrated that Chinese teams are capable of creating advanced AI solutions at significantly lower costs compared to Western companies.
According to IG market analyst Tony Sikamora, the announcement led to a loss of $314 billion in expected value for private companies OpenAI and Anthropic. Investors began reviewing the competitive dynamics of the foundational AI model market and questioned the effectiveness of US export controls on advanced chips in curbing AI progress in China.
The Chinese semiconductor ecosystem emerged as a winner. SMIC shares rose by more than 5% as increased competition in AI models is expected to stimulate further infrastructure spending. The subsequent announcement of Alibaba Qwen3.8 Max reinforced this trend, causing the ChiNext Composite index to rise by 3.6%.
Portfolio manager Gary Tan of Allspring Global Investments noted that the main beneficiaries remain companies involved in the AI infrastructure layer. China's focus on open-source AI will accelerate the adoption of domestic advanced models, which will require more computing resources and increase demand for basic hardware, especially networking equipment and memory chips.
Memory chip manufacturers are clear winners. Despite the high efficiency of the 896-expert MoE sparse architecture, Kimi K3 still requires about 1.4 TB of memory even after low-precision compression, sustaining constant demand for SK Hynix, Samsung Electronics, and Chinese memory producers.
AI agent and software developers are also benefiting, as cheaper advanced models reduce application development costs and speed up their deployment in coding, customer service, and industrial processes.
Model developers have been losers, facing a sharply deteriorating competitive environment. Zhipu AI, previously considered a leading developer of open models in China, saw its Hong Kong-listed shares drop nearly 40% over two trading days, reflecting market concerns about model-level competition. The announcement that K3 weights will be fully open on July 27th will allow enterprises to run the model independently, without reliance on cloud services, intensifying price pressure across the entire model creation value chain.
Even high-performance chip manufacturers like NVIDIA and AMD face renewed scrutiny regarding the premium pricing they set due to scarcity.
The combination of Kimi K3 and Alibaba Qwen3.8 Max over several days demonstrates that competition in advanced AI models has moved beyond US laboratories. Morningstar analyst Malik Ahmed Khan expressed skepticism about a direct causal link to the sell-off of large technology companies, arguing that American enterprises will not abandon existing cloud providers for Chinese open-source models due to security and regulatory compliance requirements. Nevertheless, the competitive pressure on model pricing and the confirmation of alternative approaches to efficiency represent sustainable changes in the AI investment thesis, requiring a finer distinction between companies benefiting from building AI infrastructure and those whose business models depend on maintaining the scarcity of advanced models.
