At a time when computing power demand is characterized by single-source data as being at an "explosive period," NVIDIA's data center business has seen quarterly computing power sales soar to 89 billion, with a year-on-year growth rate of 117%. However, GPU prices remain in a primitive stage requiring countless emails and phone calls to inquire about quotes. In this imbalance, Arthur Hayes-led Flop Labs recently announced the establishment of a GPU computing power exchange, allowing the buying and selling of computing resources on the platform, where AI agents can independently purchase computing power from miners using FLOP tokens. Almost simultaneously, Alibaba Cloud entered from another end at the application level, launching the enterprise-level collaboration platform "Boundless" on September 2, 2026, which breaks down complex tasks to multiple agents with different roles, organizing members and assets in a "project space," driving tasks forward through group chats, and transforming the produced documents and plans into reusable digital assets. This is a typical dual narrative of East and West: one end attempts to use tokens to incentivize and open up the decentralized computing power market, establishing a standardized track for upstream GPU pricing and trading, while the other relies on the platform capabilities of cloud vendors to rewrite the corporate collaboration paradigm, embedding multiple agents' capabilities into organizational processes downstream. When these two fronts are launched at the same time, stretching the boundaries of AI agents from the computing power and collaboration sides, a longer main line begins to emerge — AI agents are no longer merely isolated "smart tools," but are seen as the infrastructure of a future digital society, constructed and competed for on the levels of computing power, identity, and collaboration, just like "water, electricity, and coal."
Soaring computing power, yet still stuck in price inquiry
As enterprises begin to seriously plan AI agents as "water, electricity, and coal," the imbalance on the upstream side of computing power has reached an alarming level. NVIDIA's data center business reportedly achieved quarterly computing power sales of 89 billion, with a staggering growth rate of 117%. According to a single source, this set of numbers, repeatedly cited by Flop Labs, has almost become the standard footnote for "computing power demand entering an explosive period": any company willing to add a "meter" for the agent is simultaneously flocking to the same supplier for the same type of hardware. Computing power is being fought over like electricity in the industrial era, yet the method of buying and selling is still stuck at the "calling to inquire about prices" stage, creating a time mismatch that sets the stage for everything that follows.
Flop Labs pointed out directly in its briefing: GPU pricing still lacks a unified standard; users typically need to contact suppliers one by one for quotes, with computing power buying and selling primarily occurring through peer-to-peer communication, lacking a centralized marketplace like the energy market with public quotes and spot matching. Sellers quote prices arbitrarily based on customer relationships and conversational tactics, while buyers make investment decisions amid a fog of information — maintaining such primitive trading habits atop the substantial figure of 89 billion and a growth rate of 117% constitutes a structural contradiction in itself. The demand side has entered a new cycle of "infrastructuralization," while the supply side remains in the old world of "personal intermediaries," naturally opening a window for projects like computing power exchanges: whoever can first abstract prices, credit, and matching mechanisms into a set of industry-accepted standards will have the opportunity to master the hub for computing power flow in the future AI agent era.
Flop Labs: Facilitating GPU trading with FLOP tokens
As computing power trading remains in the era of "adding friends on WeChat and privately quoting prices," Flop Labs offers the answer by abstracting this primitive process into a programmable market: miners list their idle or for-sale GPU computing power on an exchange, and AI agents no longer rely on humans to inquire about prices or negotiate; instead, they directly purchase computing power from miners using a token called FLOP. According to a single source, the basic concept of this exchange is to allow computing power itself to be publicly listed and matched like on-chain assets, with the token playing the role of unified pricing and settlement. As for key details such as the total supply of FLOP, distribution, or staking mechanisms, the outside world is still in the dark, but this actually highlights that the project currently leans more towards a "concept launch": first capturing the narrative of standardizing and tokenizing computing power, then gradually filling in the institutional design.
A more tension-filled idea is viewing AI agents as true economic subjects — they can hold FLOP on-chain, sense computing power prices, and autonomously decide when and from which type of miner to purchase how much computing power. Consequently, the crypto economy is no longer merely a tool for human speculation, but is directly embedded in the supply side of AI infrastructure; computing power is no longer a closed commodity of cloud vendors but has been decomposed into "water, electricity, and coal" that any agent can immediately call upon. Symbolically, the leadership of Flop Labs by renowned figures in the crypto finance field like Arthur Hayes is itself a signal: when individuals who once focused on designing financial derivatives start betting on the new paradigm of "AI agents autonomously consuming computing power," it indicates that this crossover field is being incorporated into the mainstream narrative of the crypto circle, questioning who will define computing power prices and who will master the direction of computing power flow in the future, not only as a technical competition but also as a redistribution of financial discourse power.
Boundless: Assigning projects to a team of agents
In harmony with Flop Labs' betting on "letting AI buy power itself" at the computing power end, Alibaba Cloud has chosen another path at the application level: no longer fantasizing that a large model will meet all demands within the enterprise, but rather breaking down complex work for a team of agents with specific functions. "Boundless," which launched public testing on September 2, 2026, has incorporated "multi-agent collaboration" into its product positioning from the beginning: in the same project, demand analysis, data retrieval, copywriting, and data organization can be consecutively completed by different agents, while human members, acting as project managers, set goals, assign tasks, and correct directions alongside.
This collaboration is solidified in a very specific work paradigm: the project space acts as a temporary war room for the enterprise, gathering members, permissions, and related assets into the same interface; group chats become the main battlefield for humans and agents, with tasks being taken on, supplemented, and questioned in conversation threads until they form periodic outputs; all process outcomes and final deliverables are uniformly stored in an asset repository, becoming digital assets that can be directly invoked in the next round of projects. For enterprises, this means that AI agents are no longer isolated "smart plugins," but can be embedded within the entire processes of project management, knowledge management, and collaborative communication, reshaping the internal division of labor and memory structures of the organization in a team-oriented, multi-role way.
Tokenized computing power versus cloud platform collaboration, East and West place their bets
If "Boundless" is about reshaping the rules of "who works together and how to collaborate" internally within enterprises, then Flop Labs' GPU computing power exchange is about rewriting the upstream script of "where computing power comes from and at what price it is bought." According to a single source, Arthur Hayes-led Flop Labs directly places AI agents at one end of the computing power market: agents themselves hold FLOP tokens and purchase GPU time slots from miners in the computing power exchange, aiming to transform the GPU supply, which traditionally required manual price inquiries and lacked standardization, into an algorithm and token-driven automated matching process. The background is that NVIDIA's data center business has been marked for a quarterly computing power sales figure of 89 billion, with a year-on-year growth rate of 117%, yet users still need to bargain with suppliers one by one. This dissonance provides Flop Labs with a narrative path of "solving the pricing of water, electricity, and coal" first.
In contrast, Alibaba Cloud has chosen to bet on organizational collaboration rather than the source of computing power. When "Boundless" launched public testing on September 2, 2026, the product narrative did not discuss GPUs or tokens but emphasized multi-agent collaborative work: complex tasks are broken down into different responsibilities among agents, and human members chat collectively in the "project space," creating reusable digital assets from documents, decisions, and process outputs. According to the narrative suggestions of the briefing, the Western line tends to use token incentives and decentralized markets to abstract computing power into tradable resource units; the Eastern line, however, is spearheaded by cloud vendors who build platforms, treating agents as new roles in corporate project teams, with the first focus being on how humans and agents can collaborate within existing processes. There is no confirmed direct correlation between the two, but within the tension of the synchronous evolution of computing power trading standardization and enterprise collaboration paradigms, AI agents are gradually being shaped from a singular model into an ecological role that can autonomously purchase "electricity" and embed in organizational division of labor.
Once water, electricity, and coal are in place, how will the AI agent ecosystem grow?
From the upstream GPU computing power exchange to the downstream enterprise multi-agent collaboration platform, these two fronts pointed out by a single source collectively indicate a change: AI agents are no longer just "smart models," but are arranged into a complete economic and organizational structure — they need to learn to autonomously procure "electricity" in a standardized computing power market, while also learning to collaborate with humans and other agents in enterprise project spaces. This ecological role transition theoretically leaves immense imaginative space for tokenized computing power and multi-agent workflows, but currently, all key information comes from a single media channel, with data quality assessed as medium; NVIDIA's 89 billion computing power sales, Flop Labs' tokenized computing power concepts, and Alibaba Cloud's collaboration logic in "Boundless" can only be viewed as limited credible samples, not to mention that details such as the FLOP token economic model, GPU computing power exchange launch timeline, pricing model for "Boundless," and initial customer cases have yet to be disclosed. In this context, a more realistic focus from an investment and industry perspective should center on several verifiable directions: whether the tokenized computing power market can genuinely achieve different curves in price discovery, liquidity, and resource matching compared to traditional cloud vendors, whether the actual adoption of the multi-agent enterprise collaboration platform can penetrate the pilot phase and become a long-term production tool, and how the differences in regulation, corporate culture, and tech stack between the Eastern and Western infrastructure pathways will ultimately shape distinctly different AI agent ecosystems.
Join our community to discuss together and become stronger!
Exclusive Hyperliquid benefits for AiCoin: https://app.hyperliquid.xyz/join/AICOIN88
Exclusive Aster benefits for AiCoin: 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。
