From July 21 to 23, 2026, global technology and finance companies laid their cards on the table regarding AI infrastructure and business: Alphabet reported second-quarter earnings with both revenue and profit exceeding expectations, yet its stock price fell over 3% after hours, not due to insufficient growth, but because it continued to push capital expenditures into a higher range; meanwhile, OpenAI announced the commencement of at least a $20 billion data center park in Georgia, and Galaxy Digital planned to issue approximately $3.5 billion in high-yield bonds with an estimated yield of about 9% to finance a Texas project, as high-risk debt funds began to connect with computing power expansion; on the other side, ServiceNow revealed a 25% year-over-year increase in subscription revenue, and the annual contract value of its AI business had exceeded $1 billion, while Alibaba's Qwen team launched Qwen Image 3.0 with support for 4500 tokens of input, attempting to prove with tangible contracts and product iterations that the infrastructure “burned” is now generating cash flow; capital market sentiment is also simultaneously reassessing: Revolut's employee equity transfer confirmed a valuation of about $115 billion, and the U.S. House of Representatives passed the "Stop Insider Trading Act" with a vote of 232 to 198, setting new political and financial backgrounds for this AI arms race amidst the tug-of-war between high-risk preferences and fair order—in the bifurcation between computing power input and cash returns, who continues to increase investments without limits and who has begun to respond to market doubts with real cash flow has become the most critical question behind all the numbers over these three days.
Alphabet Earns More Yet Falls: Burning Money Causes Panic
In the torrent of information from these three days, Alphabet delivered a typical report showing “fundamentals are perfect, yet stock prices are declining.” In the second quarter, revenue was approximately $119.8 billion, an increase of 24% year-over-year, with the core advertising business continuing to recover; more crucially, cloud business revenue soared by 82% year-over-year, with an operating profit margin of about 34%, significantly exceeding the market's previous consensus expectations. According to textbook logic, such a combination of growth rate and profit margin should support an optimistic narrative that the "AI transformation has begun to materialize," but what truly weighed on investors' minds was another string of more eye-catching numbers.
The heaviest line item in the report was the second-quarter capital expenditures of about $44.9 billion, along with the subsequent upward revision of the full-year capital expenditure guidance. Since 2023, Alphabet, Microsoft, and Meta have alternated in ramping up data center and GPU purchases, viewed within the larger context of an “AI infrastructure arms race,” and this quarter's expenditures showed no signs of peaking but accelerated once again, causing many funds to start worrying that management's pace of investment in computing power is outpacing the market's expectations for return cycles. After hours, GOOGL's stock price fell over 3%, which was not a dramatic drop but was sufficient to indicate that divergences are widening: some investors are willing to continue tolerating high expenditures, viewing cloud profit margins as a preview of future cash flows, while others see the continuous increase in capital expenditures as a gamble without a clear endpoint. This made Alphabet's earnings report one of the turning points for current sentiment—it laid the core contradiction of the entire AI infrastructure cycle on the table: with revenue and profits climbing steadily, the capital market began to publicly question whether this long-distance race is worth continuing to sprint forward at such a money-burning pace.
OpenAI and Galaxy: High-Leverage Bet on Data Centers
At the same time investors were furrowing their brows at Alphabet's capital expenditures, OpenAI provided a completely different answer: it chose to write its computing power demands for the next few years, or even decades, directly into a massive industrial site in Georgia. The official disclosure was for a data center park of at least $20 billion, covering approximately 1,400 acres, and expected to require about 3.2 gigawatts of power supply. This is not an ordinary data center but a long-term battle revolving around computing power, energy, and land: first locking in power and space, then assuming it can turn these heavy fixed costs into a moat with a continually expanding model and applications. Unlike Alphabet, which was forced to explain at the earnings call “why it continues to spend money,” OpenAI chose to send a simpler, more straightforward signal to the market through the project size itself—it is confident that this race for computing power is far from reaching its peak.
If OpenAI is building a giant “city-state of computing power” with its own and collaborative capital, Galaxy Digital is attempting to write a similar story into its fundraising documents for high-yield bonds. It plans to issue approximately $3.5 billion in junk bonds with a target yield of about 9% to finance a data center project in Texas. Junk bonds traditionally belong to the fundraising tools of high-risk companies or projects, and are now beginning to be used to support AI infrastructure; this deal is seen as a landmark case for high-yield bonds penetrating this field. High leverage and high coupons mean extreme duality: if AI computing power demand truly continues to rise as OpenAI has wagered, the cash flows from such projects can cover expensive financing costs, bondholders take away a 9% yield, and equity holders enjoy amplified residual benefits; but if computing power demand falls short of expectations over the next few years, underloaded and low-utilization data centers will cause the high-yield bond structure to amplify the risk into a higher probability of default. There is currently no public data on whether Galaxy's bond issuance will be successful or how much it will be subscribed, but as Alphabet's stock price corrects for “burning cash,” the credit market is still willing to pay a higher coupon to continue supporting AI infrastructure, indicating that the capital surrounding this arms race is not simply turning away but is repricing the boundaries of risk and return across different asset classes.
ServiceNow and Qwen: From Stories to Cash Flow
While the capital market is stress-testing rental rates for data centers and junk bond risks, another quieter path is already providing answers—some companies are beginning to use “renewal fees” instead of “stories” to value AI. In ServiceNow's latest earnings report, subscription revenue was about $3.9 billion, a 25% increase year-over-year; this is not a one-time large deal, but rather long-term contracts and upgrade packages laid out across global corporate processes. Among these, the annual contract value of its AI business has exceeded $1 billion, meaning that the AI capabilities embedded in work order distribution, process automation, and knowledge base searches are no longer just highlight features in demonstration videos but are now written into customer budgets with clear terms and predictable cash returns, transforming into cash flow that can be audited and discounted.
On the other end of the application layer, Alibaba's Qwen team is trying to respond to the question of “Can AI monetize?” with the tool itself. Qwen Image 3.0 has elevated instruction input to 4500 tokens, which is 4.5 times that of the previous generation. The significance of this number lies not in the model being “smarter” but in its capability to handle an entire page of complex layout requirements: from copy hierarchy to color specifications, from multi-element layouts to brand constraints, approaching the length typically used in professional designers' work specifications. The Qwen team continues to iterate on image generation models, precisely targeting professional design and complex content creation scenarios, allowing businesses and creators to convert time originally spent on repeated communication and revisions into quantifiable efficiency gains. ServiceNow locks its AI capabilities into annual contracts, while Qwen uses tools to directly undertake high-value creative processes; these two models collectively point to a change: investors are starting to shift their focus from the computing power curve to renewal retention and actual usage data, and the core of AI valuation is transitioning from “the potential of the technology story” to “functions that are repeatedly paid for.”
Political and Business Trust Game: House Bill and Revolut
As investors begin to price AI stories with renewal fees and usage data, Washington is trying to redefine another boundary. The U.S. House of Representatives passed the "Stop Insider Trading Act" with 232 votes in favor and 198 against, appearing to be a technical compliance arrangement, but it genuinely addresses long-standing public distrust regarding lawmakers' stock holdings, insider trading, and the collusion of politics and business: the public no longer accepts the divided role of “both referee and player.” The bill intends to restrict lawmakers and their families from trading stocks, with the technology and AI sectors primarily included in political considerations—once legislators can no longer easily trade high-growth assets, the public pressure regarding “who profits from the AI boom” may shift from individual transactions to institutional design. However, the House is just the starting point, and the subsequent timetable for Senate review remains unclear; whether the bill can truly be enacted is still uncertain, and this uncertainty itself will be viewed by the market as a pending institutional risk.
On the other end, the capital's obsession with growth narratives has not cooled. Revolut confirmed a valuation of about $115 billion in its employee stock transfer transaction, with a share price of about $2017, and amidst increasingly heated regulatory discussions, private equity and secondary markets are still assigning high-risk preference prices to fintech and growth tech companies, a stark contrast that is highly symbolic: the public calls for constraints on the relationship between power and the market, while capital is willing to pay a premium for “the next phase of financial and technological infrastructure.” From the perspective of the trust and valuation game between politics and business, pricing for AI and tech assets has never been solely about performance and stories; one end is tied to financial statements, contracts, and product usage data, while the other is tied to institutional trends and public sentiment, ultimately determining who will foot the bill for this arms race depends on where the market and politics redefine acceptable boundaries of risk and return.
The Next Act in the AI Arms Race: Who Can Survive Under Capital Constraints
Since 2023, companies like Alphabet, OpenAI, and Galaxy have continued to raise the threshold for investments in data centers and GPUs: Alphabet's second-quarter capital expenditures soared, yet its stock price fell after the earnings report; OpenAI wagered at least $20 billion on its Georgia park, and Galaxy attempted to leverage about $3.5 billion in junk bonds with a yield of about 9% to boost its Texas project, representing a typical "capital-intensive + high-leverage" computing power track. On the other end, ServiceNow's subscription revenue is steadily growing, its AI business annual contract value exceeding $1 billion, while Qwen uses Qwen Image 3.0 to align with more complex real-world scenarios, following an “industrial product + contract cash flow” application path. Investor sentiment has thus been torn: concerns over the former regarding return cycles and leverage risks, yet willing to pay valuation premiums for the latter stories that have already produced cash flow. As capital costs rise and external constraints from regulation and public sentiment strengthen, the focus of this arms race is shifting from “who has the highest computing power” to “who has more stable cash flow and controllable risks.” However, the outcome of Galaxy's bond issuance is yet to be determined, and the "Stop Insider Trading Act" still needs to go through the Senate process. Before these key variables are settled, any judgments about short-term price trends or legislative directions can only be viewed as tests in a state of high uncertainty.
Join our community to discuss and grow stronger together!
AiCoin's exclusive Hyperliquid benefits: https://app.hyperliquid.xyz/join/AICOIN88
AiCoin's exclusive Aster benefits: https://www.asterdex.com/en-US/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。
