On June 9, 2026, two seemingly unrelated yet actually parallel news pieces surfaced on the same timeline: on one hand, the Financial Times reported that Anthropic finalized a private credit deal of about $35 billion with Apollo and Blackstone for its "growth plan" and chip project, pulling this story, originally tied to infrastructure and financial engineering, into the high capital density battlefield of computing power; on the other hand, Beating disclosed that OpenAI co-founder and CEO Sam Altman and Chief Scientist Jakub Pachocki jointly released a long-term roadmap, announcing the company's entry into a third development stage centered on "ubiquity and safety," emphasizing that "the ultimate goal of artificial intelligence is to serve humanity and decentralize power, rather than replace human decision-making," and proposing the creation of "automated AI researchers," envisioning that by 2028 AI will take on most of the work in its own research and development. On the funding side, this high-leverage private credit is being funneled directly into chip and computing power infrastructure; on the technology side, the research and development system is beginning to plan for models to accelerate their own evolution. These two leads point to leading AI companies accelerating their independence in computing power foundations and research paradigms. It is important to note that whether regarding the key terms of this $35 billion credit or the specific technological path in OpenAI's roadmap, publicly available information is highly limited. The data and quotes in the briefing primarily come from single-source reporting. Nevertheless, even beneath these incomplete pieces, a clear outline begins to emerge: the AI industry is transitioning from simple "money-burning expansion" to a new stage of "high-leverage capital-driven infrastructure investment plus AI core self-evolving research and development systems."
$35 Billion Credit Line: Anthropic's Bet on Chips
In this migration towards "high capital density infrastructure," the most striking number comes from Anthropic. According to the Financial Times, Apollo and Blackstone have reached a private credit arrangement of about $35 billion with this leading AI company to support its "growth plan" and "chip project." In the technology industry, especially among unlisted AI companies, such a scale of private credit is extremely rare, resembling financing scales typically seen in traditional infrastructure or energy projects, clearly indicating that AI chips and computing power foundations have entered a stage requiring massive upfront capital investment.
More crucially, this money is clearly aimed at two areas: one is overall expansion, and the other is the chip project, essentially paying for computing power autonomy. For Anthropic, continuously increasing its own or controllable chip and computing power infrastructure is an attempt to shift from being a "buyer of computing power" to a player with greater initiative in the technology stack, thereby gaining more pricing power over future model training and inference costs. The decision to complete this increase in collaboration with Apollo and Blackstone through private credit rather than traditional public market financing reveals another issue: both borrowers and lenders are willing to leverage greater risks to bet on who will control the upstream power of the value chain in the next phase of AI competition.
Private Credit Takes the Stage: Wall Street Bets on AI Foundations
Signing opposite Anthropic this time are Apollo and Blackstone, two alternative asset management giants that have long been deep into private credit. They are not short-term financial investors chasing trends, but rather "debt-oriented players" accustomed to viewing cash flow and asset safety through a decade-long lens. When such funds choose to open a line of credit of about $35 billion aimed at Anthropic's "growth plan" and "chip project," the intention is clear: a group of individuals on Wall Street, skilled in calculating risk-return ratios, is using their most familiar tools to bet on one thing—future demand for chips and computing power infrastructure will not be a fleeting phenomenon but a long-term cash flow curve.
In recent years, private credit has quietly undergone a role transformation in the tech circle: from supplementary financing to a key channel for tech giants to secure large amounts of financing, especially in the capital-intensive infrastructure investment space. The public market is becoming increasingly picky about the "burning money to buy servers" narrative, while private credit has filled this funding gap with higher pricing and more complex terms. The $35 billion that Anthropic received can almost be seen as a concentrated representation of this trend in the AI sector—money is no longer just chasing hot products on the application end but is directly being poured into the most substantial assets at the AI infrastructure level.
However, in an AI track where both the business model and cash flow remain highly uncertain, supporting expansion with huge debt itself is a deliberately magnified leverage. For Anthropic, if the chip project and computing power layout truly lead to monopolistic technology and market positions, the fixed costs of debt could be easily diluted by high profit margins, amplifying equity returns; conversely, if the industry competition landscape, regulatory environment, or technological path deviates, this debt-centric financing structure could rapidly backfire, accelerating cash flow pressure. Especially with the key terms of interest rates, collateral arrangements, and repayment periods all undisclosed, it is nearly impossible for outsiders to estimate how "heavy" this money is on the balance sheet, only knowing that Wall Street and leading AI companies are using the same leverage to push the potential benefits and systemic risks of AI infrastructure to a higher and more difficult-to-reverse level.
OpenAI's Third Stage: The Trade-offs of Ubiquity and Safety
On the same day that Anthropic leveraged its investments in computing power facilities, Sam Altman and Chief Scientist Jakub Pachocki stepped forward to outline a seemingly more "restrained" roadmap for OpenAI: the company announced its entry into a third development stage centered on "ubiquity and safety." Compared to the earlier narrative that nearly solely pursued breakthroughs in model capabilities and user scale, this roadmap deliberately downplayed the tone of "bigger, faster," repeatedly stressing that "the ultimate goal of artificial intelligence is to serve humanity and decentralize power, rather than replace human decision-making," attempting to directly address public anxiety regarding the concentration of power in large models and even the thoughts of "replacing society's choices."
Yet, within this more tempered declaration lies a more radical thread: OpenAI has formally included "automated AI researchers" in its long-term planning, providing a rough timeline vision—by 2028, AI will take on most of the work in its own R&D. In other words, those responsible for accelerating the iteration of the next generation of models will no longer just be human engineers but the automated researchers derived from existing systems. The temptation for efficiency is evident: once the research and development pipeline is highly automated, model version upgrades can proceed at a pace far exceeding human capability; however, the roadmap does not provide specific technological paths, stage goals, or design details for safety gates. How humans can maintain ultimate control in a system driven by AI-led research and development remains deliberately left blank, making it the most crucial and difficult-to-quantify variable in this "ubiquity and safety" game.
AI Researchers on Duty? The Tug-of-war over Human Control
If the goal of "automated AI researchers" is even partly realized, the power structure of the research and development chain will be rewritten: from proposing hypotheses and designing architectures to generating experimental plans and tuning parameters, an increasing number of key links will be completed by the models themselves, moving human experts from "hands-on model creators" to reviewers and gatekeepers. OpenAI even specified in the roadmap its expectations: by 2028, AI will take on most of the work in its own R&D, yet did not provide corresponding technical paths or safety constraint details, which indicates that future model capabilities will largely depend on how the previous generation of models "train the next generation," rather than being derived from human researchers painstakingly dissecting each line.
Consequently, anxieties surrounding "who makes the decisions" are inevitably magnified: when AI is both a tool and gradually becomes the protagonist of the R&D process, how much final decision-making authority can humans still hold? OpenAI repeatedly emphasizes that "the ultimate goal of artificial intelligence is to serve humanity and decentralize power, rather than replace human decision-making," framing boundaries for this self-evolution route. However, it remains silent on how "automated AI researchers" will be constrained by safety mechanisms, how they will be reviewed, or whether external regulations will be introduced. The lack of visible technical and institutional gates makes this assertion more like a preemptive "verbal firewall," and the real question that needs to be addressed is whether humans can still halt the chain reactions initiated when AI researchers begin to make daily technical and resource decisions for humanity.
From Money-Burning Expansion to the Turning Point of Self-Evolution
Looking at the same timeline: on one side, according to the Financial Times, Anthropic reached a private credit agreement of about $35 billion with Apollo and Blackstone to "pre-exhaust" future investments for chips and computing power infrastructure; on the other side, Beating disclosed that OpenAI has included "automated AI researchers" in its core narrative within its long-term roadmap, attempting to let AI take on most of its own R&D by around 2028. This is a joint action from leading companies seeking both capital autonomy and technological independence—the former locks more capital into its own computing power foundation through high leverage, while the latter attempts to control a faster technological iteration pace with fewer human resources. Consequently, the industry is turning from an externally driven model of "human-led research and development plus continuous capital infusion" to an endogenous evolutionary path of "high capital density computing power investment plus AI self-optimization": computing power debt is replacing part of equity financing, and automated researchers are diluting the decision-making power of frontline engineers. The issue, however, is that whether regarding the interest rates, collateral, and terms of this private credit, the specific allocation of funds in the chip project, or the technical details and stage goals of OpenAI's roadmap, currently, all are undisclosed. The relevant information primarily originates from single sources such as the Financial Times and Beating, significantly amplifying the interpretative space due to the opacity. In the future, with overlapping leverage from computing power debt and self-R&D, how to establish regulatory boundaries, assess potential systemic risks, and redistribute gains and losses between tech companies, funding providers, and the wider public will no longer be just a moral questioning, but will become the primary battleground that the market and policy must confront.
Join our community to discuss together and grow stronger!
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。
