〔Day Digest〕AI Capital Spending Surges: Tencent, Samsung, and SpaceX Pour Billions into Compute

Corporate AI Investments Accelerate

Tencent reported strong Q2 results with revenue of ¥204.79 billion, up 11% year-on-year, and Non-IFRS operating profit of ¥75.64 billion, up 9% ^[1]. Excluding new AI product impacts, operating profit growth was 19% ^[1]. The company's AI spending surged: R&D expenses grew 35% to ¥27.28 billion, and capital expenditures jumped 176% to ¥52.78 billion, with CEO Ma Huateng emphasizing a massive increase in compute procurement to convert model usage into revenue ^[1].

In another major move, Samsung and SK Hynix are reportedly preparing record shareholder returns exceeding 200 trillion won ($141.2 billion), potentially announced by end of August ^[2]. Samsung's returns could exceed 100 trillion won, driven by free cash flow growth ^[2]. This comes amid an AI-driven memory supercycle, with shares rising about 6% on the reports ^[2].

Hardware and Chip Innovation

Samsung Electronics has integrated Anthropic's Claude into its high-end chip development, compressing verification time from over a month to two days—a 15x efficiency gain ^[3]. The system LSI division also saw a junior engineer complete USB model development in one day instead of a month ^[3].

Synopsys announced the world's first HBM4 IP test chip validation, achieving stable operation at 9.2Gbps, the highest data rate supported by current HBM DRAM ^[4].

Intel CEO Lip-Bu Tan revealed a new memory architecture initiative, hiring SK Hynix's Shil Lee, as the company looks to return to its memory roots ^[5].

Funding and Market Moves

Shanghai Sigmarized (硕贝德) announced a private placement of up to ¥1.0958 billion for server cooling modules, smart glasses RF components, low-orbit satellite antennas, and working capital ^[6]. China's central bank pledged PSL quotas to support AI, digital economy, and urban renewal projects ^[7].

But Dan's overseas fund cleared Apple, Tesla, and other positions, adding AI hardware stocks including Intel, AMD, Broadcom, and Marvell ^[8]. AMD trades at 63x forward earnings while Nvidia trades at 24x, with historical analysis suggesting a better buy ^[9][10].

AI-Driven Energy and Infrastructure

The IEA highlighted a paradox: AI is becoming more energy efficient while data centre electricity demand surges ^[11]. Chinese state media emphasized that surging computing power demands more electricity ^[12]. Meanwhile, visions for AI infrastructure expand to the sea: Panthalassa is raising $225 million at a ~$2 billion valuation to deploy GPU nodes on ocean buoys ^[13].

SpaceX's Elon Musk stated AI revenue could surpass all other business revenue by September and eventually account for 99% of company value within 4-5 years ^[14]. The company targets 10 gigawatts of AI compute capacity by end of 2026, potentially generating $300-500 billion in annual revenue ^[14].

Policy and Industry Debates

Anthropic will apply global watermarking to Claude models post-August 2, citing EU AI Act transparency obligations ^[15]. The EU's interoperability requirements for watermark detection tools remain a key challenge ^[15]. A Guardian op-ed argues the US should nationalize OpenAI and Anthropic if markets reject them ^[16][17]. However, reports show AI has not destroyed jobs as predicted, though job roles are changing ^[18].