In the summer of 2026, a string of numbers in the capital ledger pushed the AI competition to a new scale: on one side, the South Korean government plans to inject about 20 trillion won (approximately 139 billion USD) into the KIC, marking the first time this sovereign fund is allowed to invest in domestic AI and data centers, directly writing computing power into "national assets"; on the other side, the four major tech giants set their capital expenditure cap for 2026 at approximately 745 billion USD, with cloud data centers and GPUs becoming the most expensive and most bet-on line on their profit statements. On the surface, AWS reported a revenue of 42.2 billion USD in Q2 2026, a year-on-year increase of 37%, while Azure and Google Cloud proved with faster growth rates that cloud business is the first battlefield to realize revenue in this wave, but on a longer time axis, capital is betting on distinctly different paths: Cook calls "hybrid AI" Apple's competitive weapon, hoping to drive iCloud subscriptions through a combination of terminals and the cloud, while ByteDance uses Seedance 2.5, Doubao with 324 million monthly active users, and Volcano Engine with about 16% market share to compress the profits of Douyin/TikTok onto a full-stack model to the cloud; parallel to this, Michael Saylor reviews STRC's history of returning to a $90 valuation about 70 trading days post-IPO, viewing the adjustment of about 40 trading days since it fell below the 99-100 USD range on May 28 as a new anchoring round, even pointing to a target date around September 8. Thus, within the same time window, sovereign funds, tech giants, and digital asset narratives are pulled into the same balance sheet: whether a hundred-billion-dollar hardware investment can be covered by commercial returns and digested by market sentiment becomes the true dividing line in this AI capital race.
National Team Entry: South Korea's KIC 20 Trillion Bet on AI Infrastructure
Within the same time window, the South Korean government announced plans to inject about 20 trillion won (approximately 139 billion USD) into KIC specifically for AI and data centers, which was interpreted as a direct signal that computing power and data centers are no longer just corporate capital expenditures but have been elevated to national strategic assets. More importantly, this is the first time KIC has extended investment rights to domestic assets, and this institutional relaxation itself is a clear signal from the policy level that "AI infrastructure should be backed by the national team"—sovereign funds are no longer content to be passive allocators of global assets but are willing to take on the role of builders of their own national computing power base.
From the moment KIC can invest in local assets, the South Korean data center and AI computing ecosystem receives an additional layer of implicit guarantee from national credibility: in the future, whether it's newly built parks, expanded racks, or introducing high-power servers, as long as it can be packaged into the narrative of "AI and data centers," there is a chance to connect with this long-term patient capital level of 20 trillion won. Against the backdrop of the four major tech giants setting a combined capital expenditure cap of about 745 billion USD for 2026 and continuously pouring funds into computing power and data centers, KIC's shift is also seen as a significant repricing of sovereign fund styles—from a preference for traditional financial assets with predictable cash flows to tolerating higher volatility in exchange for potential premiums on strategic tech assets. The specific timeline and targets have not yet been disclosed, but the direction is clear: in this AI capital race, whoever first ensures that the nation's sovereign funds stand behind data centers and fibers has clearly bet their domestic risk appetite on computing power and infrastructure.
745 Billion USD Arms Race: Cloud Giants Accelerate Spending
If sovereign funds have just pushed chips towards data centers, then on the other side of the table, the four major tech giants have already piled their cash flows for the next few years into a mountain—according to a single source, their combined capital expenditure cap for 2026 totals about 745 billion USD, almost transparently directed at AI, computing power, and data centers. This isn't just an annual "renovation fee"; it resembles an arms race without a ceasefire timetable: each company knows that failing to invest risks being left behind, but no one can clearly state how much of their expenditures will be written off as failed racks and obsolete chips years from now. More importantly, how these funds are allocated among the four and the specific projects they are being poured into has not been disclosed; what the market can see is just a steeply rising capital expenditure curve.
Visible returns are currently concentrated in cloud business—meaning AI computing power has first been packaged into cloud billing. Amazon's overall revenue reached 200.6 billion USD in Q2 2026, a year-on-year increase of 20%, with operating profit soaring to 27.5 billion USD, up 43% from the previous year; AWS specifically recorded a quarterly revenue of 42.2 billion USD, a 37% increase year-on-year, the fastest growth in the past 18 quarters, with operating profit reaching 16.6 billion USD, up 64% year-on-year. Comparisons from single sources show Azure's growth rate during the same period was about +43%, while Google Cloud reported an impressive +82% year-on-year growth. The steepness of the cloud revenue curve has provided a temporary comfort to capital markets: at least for now, there are buyers for these high-priced computing powers. But immediate questions arise—when the 745 billion USD in expenses continue to convert to depreciation and operational costs in the coming years, and the high growth of cloud business inevitably slows, how long can the current profit margins sustain, and how long can the existing pricing systems rise? Under the narrative of "limited returns," the market's real anxiety isn't whether this AI frenzy can bring income, but rather how this arms race-style investment will ultimately compress the return rates to a new normal.
Cook's Hybrid AI Bet: Light Assets Against Heavy Investment
When the four major tech giants set a combined capital expenditure cap of approximately 745 billion USD, charging recklessly towards data centers and computing power, Apple presented a completely different route in the first half of 2026. Tim Cook publicly dubbed "hybrid AI" as Apple's "competitive weapon"—not competing on who has larger data centers or higher GPU stacks, but rather pressing computing power as much as possible to the terminal side, completing more inferences on-device, and only throwing requests back to the cloud when necessary. Apple's layout in AI infrastructure deliberately ties terminal hardware and cloud services together, not abandoning cloud capabilities while avoiding a mere accumulation of heavy assets in computing power, contrasting sharply with the path of solely expanding cloud data centers.
What Cook is really betting on is the cash flow model behind this hybrid architecture. He emphasizes hoping to leverage AI capabilities to drive further growth in subscription businesses like iCloud, allowing users to continuously pay for "smarter devices" and "more personalized cloud services," rather than pushing the company's CAPEX to its limits in the new round of an arms race. This relatively restrained capital expenditure strategy does alleviate some concerns under the current anxiety of a "new normal for return rates"—Apple can convert more uncertainty into predictable subscription revenue, rather than gradually amortizing costs in depreciation tables years down the line. However, this seemingly "light asset" path carries risks as well: should the hybrid experience of terminal and cloud fail to distinguish itself from competitors, or depend on cooperative cloud services leading to limited cost flexibility, Apple could fall behind in both computing scale and AI service depth. Whether hybrid AI turns out to be a deft defense or a hesitant offense will be forced to face scrutiny in the next round of AI investment cycle switches.
ByteDance ALL-IN: From Seedance to Doubao
If Apple is attempting to make AI a "light asset business," ByteDance is taking almost the opposite approach—using the cash flows earned from Douyin and TikTok to build a full-stack railway from models to applications to enterprise services. Seedance 2.5, the video generation model showcased in the first half of 2026, is the most noticeable carriage on this railway: the front end features familiar short video content scenes, while the backend is pushed forward by a research team of about 2,000 people from Seed, driving model iterations. ByteDance is not rushing to package Seedance as an independent product for sale, but embedding it within its content ecosystem and creative tools, letting the model first run through commercial loops in its "own traffic pool" before outputting capabilities externally.
The entry layer is carried by Doubao, shouldering traffic and cognition. Single-source statistics indicate that Doubao has about 324 million monthly active users, placing ByteDance in a position of strength at the starting line for large models, directly competing at the scale of a medium to large social product. Unlike Apple's "invisible entrance" of stuffing AI into iPhones and iCloud, Doubao is designed as a dialogue entry and workflow hub that users clearly perceive, serving both C-end users and testing the waters for subsequent enterprise products. At a deeper infrastructure level, it is entrusted to Volcano Engine to support—one source states that Volcano Engine holds about 16% of the AI cloud market share in China, ranking second, which gives ByteDance a rare "dual identity" in China's AI infrastructure track: both a content platform and a cloud service provider. Compared to overseas cloud giants growing AI services upward from infrastructure, Chinese tech giants are largely digging downwards for infrastructure from the top of traffic and content. The share of Volcano Engine proves that this "reverse growth" path is currently viable, but the real test is whether the profits and traffic brought by Douyin and Doubao can sustain ByteDance through the entire closed-loop in this heavy asset AI infrastructure race.
STRC Anchoring Countdown: Pressure Test for Digital Asset Narrative
When ByteDance uses cash flow to build data centers, capital markets are counting returns from this AI competition in another way—the anchoring of STRC. Strategy embedded STRC into its digital asset strategy not just as a financing tool but packaged it as a "note" betting on AI infrastructure dividends. However, the real challenge lies not in the white paper but on the timeline: STRC's IPO priced at 90 USD, and it took about 70 trading days post-IPO to return to near par value; this turbulence was viewed by Michael Saylor as a volatility test that has already been passed and became a reference for him to judge price repair rhythms thereafter.
This time, starting from the drop below the 99-100 USD range on May 28, 2026, the adjustment has taken about 40 trading days, and Saylor openly replicated the memory of that "70-day return to par value," targeting a new anchoring date around September 8, marking the price fluctuation of a digital asset as a countdown exam. As the capital expenditure caps of the four major tech giants rise to about 745 billion USD, with computing power and data centers regarded as national and corporate "heavy assets," products like STRC that overlay AI narratives with digital asset exteriors are being repriced by the market: investors need to judge not only whether Strategy can make money in this AI infrastructure arms race but also whether this digital asset packaging is worthy of the risk premium for future cash flows. This anchoring process centered on September 8 will ultimately be written into the sample library of "AI narrative + digital asset" structures, serving as key evidence to support or challenge such products.
Join our community to discuss and grow stronger together!
AiCoin exclusive Hyperliquid benefits: https://app.hyperliquid.xyz/join/AICOIN88
AiCoin exclusive Aster benefits: https://www.asterdex.com/zh-CN/referral/9C50e2
On-chain Telegram community: https://t.me/AiCoinWhaleData
On-chain community: https://www.aicoin.com/link/chat?cid=N6OVMor5g
AiCoin on-chain Twitter: https://x.com/aicoinwhaledata
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
