On April 25, 2026, within the same timestamp, two curves that should have been parallel suddenly intersected: on one side is the AI track, flooded with capital and resumes, and on the other side is the on-chain world torn apart again by black swans. Public information revealed that DeepSeek is planning a new round of financing with an estimated valuation of about $20 billion, one reason being that core researchers are continuously poached by companies like ByteDance, Tencent, Xiaomi, and Yuanrong Qihang; almost at the same time, Purrlend was hacked on the emerging networks of HyperEVM and MegaETH, with a total loss on-chain of about $1,522,037.82, while Drift, which experienced an attack of over $285 million a few weeks ago, provided a vague window of "optimizing code over a few weeks and planning to restart the exchange after a fork in May to June," according to product manager Minh Don on Discord.

Looking back over the past few months from this day, two narratives are rapidly diverging. On the AI side, from Moonshot AI's valuation quadrupling in just three months, to Zhizhu Technology and MiniMax listing on the Hong Kong Stock Exchange, leading companies have used rounds of financing and IPOs to turn "valuation acceleration" into the industry's default backdrop; the $20 billion rumor of DeepSeek is just the latest coordinate on this curve. Capital and talent are fast-forwarded as if a button has been pressed, building a new power map around computing power, models, and application layers, scrambling for every name that can write papers or release model cards.

The curve of the on-chain world, however, appears particularly rough. Purrlend was drained of about $1,197,488.33 on HyperEVM, including about 449,683 USDC and other assets, while losses on the MegaETH network are estimated at about $324,549.49 based on a single source, with the specific figures still to be verified; the over $285 million hole at Drift remains, and various claims regarding external support and partners still lack multiple corroborations, relying solely on keywords like "code optimization" and "fork restart" to maintain narrative coherence. Emerging networks chase performance, yet audit coverage, monitoring facilities, and emergency experience are clearly lagging; every dollar stolen serves as a reminder: this is not a coincidence of a single protocol, but a structural weakness of the entire track.

Thus, on one side is the AI—where "money and people are still not enough"—with DeepSeek needing financing to offset talent loss, while the market is more concerned about whether it can join the next unicorn story worth hundreds of billions; on the other side is DeFi—where "money can be whisked away as soon as it comes in"—with Purrlend stepping on a landmine in the early ecosystem and Drift attempting to prove that "technological rebirth" is still feasible with forks and restarts. Funds and talent are rapidly flowing into AI, yet on-chain security is repeatedly questioned by successive hacker incidents, restart announcements, and unfinished compensation plans; this comparison forms the most glaring technology capital picture of spring 2026.

$20 billion valuation: DeepSeek's talent defense battle

While the on-chain world is arguing about the order of compensation for lost millions, the AI track on the other side has raised the stakes directly to the tens of billions range. The new round of financing that DeepSeek is planning has a market valuation rumored to be around $20 billion, with specific amounts and investors yet to be confirmed, but the number alone is sufficient to send a signal: on the Chinese local model table, DeepSeek is considered one of the few players that "must be pushed to this price point."

This is not an isolated event, but a continuation of a series of accelerated valuations over the past few months. In the first three months of 2026, Moonshot AI completed three consecutive financing rounds, with a solitary source claiming its valuation has quadrupled compared to the end of the previous year; almost on the same timeline, Zhizhu Technology and MiniMax listed on the Hong Kong Stock Exchange, directly moving the story of large model companies into the public capital market. When viewed together, it is hard to explain this with "local bubbles"—the Chinese AI track is rapidly entering a stage of normalized high valuations: unlisted companies align with the pricing of financing samples like Moonshot AI, while listed companies validate the narrative with stock prices and market capitalization.

In this context, DeepSeek's choice to propose a $20 billion valuation at this time is not just about telling a bigger story. Another motivation mentioned by several traders is more combustible: recently, DeepSeek has seen core talent loss, with a number of researchers poached by ByteDance, Tencent, Xiaomi, and Yuanrong Qihang. For a large model company, this means the most important asset is not algorithm papers but those dozens of people who can continuously produce papers and engineering results; when they start to be systematically "packed away," management must use a more adequate equity pool and salary budget to stop the trend of loss, which often relies on a new round of financing to buttress.

In other words, the $20 billion valuation is not only the capital's pricing of DeepSeek but also the "quote" in this round of talent warfare. On one side are internet giants using mature cash flows and platform resources to siphon off experienced researchers; on the other side are new forces, from autonomous driving to multimodal applications, making offers with more aggressive salary promises and project discourse power. Caught in the middle are companies like DeepSeek, seen as "next-tier leaders"—to retain talent in a competitive environment where Moonshot AI's valuation surged fourfold and Zhizhu Technology and MiniMax have already listed, they must raise their own valuation to a sufficiently high range to maintain negotiating power in terms of options and future space.

Ironically, the market is circulating conflicting versions about how much money DeepSeek will secure in this round of financing and who the investors will be, with no authoritative confirmations yet, but almost all rumors assume one point: regardless of the scale of funds, the core purpose revolves around "seizing and retaining talent." From the end of 2025 to the beginning of 2026, the continuous financing and public offerings of leading AI companies quickly raised salaries and options expectations across the entire sector; by the spring of 2026, DeepSeek's valuation of $20 billion appears like the latest benchmark, etching this new normal of "high valuations + fierce talent grab" onto the timeline.

Purrlend hacked for $1.52 million: The first test of new chain security

In the same week that capital was inflating prices for "people" in the AI track, the on-chain world on the other end was put on risk alert by "code." Around April 25, 2026, the multi-chain lending protocol Purrlend was confirmed to have been attacked on the emerging networks of HyperEVM and MegaETH, with a "static yield" originally written into the smart contract turning into a dynamic loss at the level of $1.52 million within minutes.

On-chain data presents a stark numerical picture: this attack resulted in a total loss of approximately $1,522,037.82, of which the HyperEVM network was the main battlefield, with a single chain losing about $1,197,488.33, including about 449,683 USDC; the rest was scattered across various tokens. The losses on MegaETH were relatively smaller, but still at the "hundred-thousand-dollar level," with a single tweet estimating around $324,549.49, a figure still pending validation.

To the average user, Purrlend is just "another multi-chain lending protocol"; but to the attacker, it’s a hunting ground misaligned across multiple emerging networks. Purrlend chose to deploy business synchronously on new networks like HyperEVM and MegaETH to achieve higher capital efficiency and discourse power—this is about seizing incremental market share for the protocol; for attackers, it provides a larger opportunity created by multiple battlefronts: as long as a gap is torn open on one defensive line, the capital routing and risk control logic across other networks will also be forced to bear the brunt.

Currently, public information has not disclosed the specific types of technical vulnerabilities, and there is no authoritative conclusion on the identity of the attacker; circulating claims such as "insider involvement" also lack confirmation at the judicial level. However, even in the absence of technical details, the distribution of assets has already painted a clear outline:
● HyperEVM serves as the "amplifier"—with a loss reaching seven figures, compounded by nearly 450,000 USDC, making this chain the most concentrated risk part;
● MegaETH, meanwhile, fulfills the "extended front" function—losing hundreds of thousands of dollars, not as much in absolute terms as HyperEVM, but it means attackers can continue to maneuver chips, split paths, and dilute tracking difficulties on another emerging network.

This incident also marks the first clear cost of security for emerging networks in 2026. Networks like MegaETH and HyperEVM are still in the ecological early stage, performance metrics are repeatedly emphasized, but security audit coverage, on-chain monitoring infrastructure, and community emergency experience are significantly thinner compared to mature public chains that have undergone multiple rounds of attacks and iterations. Multi-chain DeFi protocols like Purrlend, while chasing the traffic dividends of new networks, expose themselves to areas where audit rhythms, monitoring tools, and cross-chain governance have yet to keep up.

In numerical terms, this attack is merely a "hundred-thousand-dollar event"; but structurally, it represents a systemic trial following the combination of new public chains and multi-chain protocols:
● Multi-chain deployment makes protocols look "better" on TVL, but simultaneously bet on different security cultures and varying infrastructure maturity;
● Once any single chain's contracts, parameters, or integrations show weak links, attackers can complete "low-cost breakthroughs" on that chain, leveraging liquidity and asset compositions from other chains to amplify profits and diversify risks.

The Purrlend incident thus becomes a clear watershed: on new networks like HyperEVM and MegaETH, security is no longer a course to be made up "once the ecology is ready," but rather a prerequisite that all DeFi protocols must write into their models, calculate costs, and pay for before choosing multi-chain expansion. The $1.52 million loss is just the first bill of this new phase.

After the $285 million heist: Drift bets on rebirth within weeks

If Purrlend's $1.52 million is just the first bill of the new phase, then what Drift encountered is a boldfaced hole on the entire industry's ledger—according to a single source, the attack event resulting in losses exceeding $285 million directly placed one of the leading derivative trading protocols under "public execution" on the chain.

Before the incident, Drift represented one end of a "mature derivatives market": contract depth, trading tools, market-making ecology, all viewed by many traders as one of the infrastructures on their public chain. After the attack, the core asset pool of the protocol was instantly drained, and while the label of a leading derivative does not change, what has changed behind this label is that it has become one of the most glaring safety failure examples in the DeFi world in recent years.

There is a circulating claim in the community that the attack occurred "on April 1," but this currently only appears in a single tweet, lacking cross-verification from multiple reports and official announcements. While the timing remains unresolved, the ledger hole is exceptionally clear: a gap of hundreds of millions of dollars has reset an entire risk control narrative, sweeping numerous high-leverage traders' positions and trust into the ruins.

After weeks of silence, a signal of change appeared on the official Discord. Product manager Minh Don surfaced around April 25, presenting the team's "rebirth script": first optimizing and restructuring the codebase over several weeks, then, on that basis, launching a re-online version after a fork. He didn't discuss grand visions but rather compressed the story into a technical project: to dismantle, rewrite, and then reattach the problematic system back to the main network.

According to Minh Don, the rough time window for the restart is marked in May or June 2026. In other words, from now on, Drift has a countdown of "weeks" for a cold start—completing repairs, testing, and deployment on the code level within weeks, and then putting the plaque back up in a market sentiment that has not fully calmed.

However, beyond the timetable, the real key determining whether this restart can occur lies almost entirely in the shadows of "unpublished" information. Minh Don has not yet presented a clear user compensation plan: regarding the funds involved in the losses, at what time point, what discount, and with what assets they will be settled, and what proportion the protocol team is willing and able to bear, all remain listed as "to be announced."

Equally unresolved is the truth behind external blood transfusions. There are rumors suggesting that Drift may have secured around $147.5 million in external support, even naming potential partner organizations, but this information currently only appears in single-source tweets, lacking official statements and cross-confirmation from multiple media, and remains unverified. For protocols that have experienced hacker attacks, such news of "financial backers on the way," if not quickly solidified, can easily evolve from easing expectations into new sources of distrust.

Historical experience is neither optimistic nor despairing—over the past few years, multiple instances of "coming back after being blown up" have occurred in the DeFi world: some have slowly pulled liquidity and users back through forks, code restructures, and governance restarts; while others have completely lost credibility in the second or third vulnerabilities, becoming footnotes in textbooks. The line of success or failure often depends on two factors: the thickness of security audits and whether there are communities and financial players willing to step up and provide liquidity during the coldest times.

Drift is now betting on a highly compressed window: delivering a safer set of code, a new exchange after a fork, and a widely accepted compensation and governance arrangement within a "few weeks." The problem is that at this moment, where compensation mechanisms have yet to be announced and external support remains speculative, will those old users with breached positions who watched their funds evaporate still be willing to hand their margin over to the same developers and the same logic? This is a trust test that any technological update cannot bypass.

Capital chasing AI, hackers targeting DeFi: The collision of two narratives

On the same timeline, two worlds are accelerating in opposite directions. On one side of the screen, Moonshot AI completed three consecutive rounds in the first three months of 2026, with its valuation more than quadrupling compared to the end of the previous year; Zhizhu Technology and MiniMax listing on the Hong Kong Stock Exchange told the story of "models, computing power, and commercialization" growth. The latest to step into the spotlight is DeepSeek, rumored to be planning a new round of financing with an estimated valuation of about $20 billion—despite the specific amount, investors, and timeline still being undecided, the market is already impatient to bring it into the "high valuation camp."

On the other side of the screen, there's Drift, leaving behind a black hole exceeding $285 million (according to a single source), and Purrlend having $1,522,037.82 in assets whisked away across two emerging networks. Drift's old users are waiting in Discord for a compensation plan that has yet to take shape, while Purrlend's lenders quietly compare the loss details against a long string of transaction records on the blockchain—here there are no high valuations, only the question of who will bear the cost.

Capital and people are skewing toward AI. The reason DeepSeek is forced to consider high valuation financing at this time is straightforward according to traders: core researchers are being poached by ByteDance, Tencent, Xiaomi, and Yuanrong Qihang. This detail illustrates not just the internal competition between AI companies, but a larger reality—among scarce skills in algorithms, security, and underlying systems, the time of the very top talent is locked in by the AI track with premium salaries and options.

Security audit firms, underlying protocol engineers, and those doing formal verification and monitoring systems are already in short supply. Now, AI projects with stronger payment capabilities and more "attractive" valuations are simultaneously queuing with on-chain protocols to grab the same batch of talent and resources. The result is that DeFi's attempt to conduct a thorough secondary audit and establish a complete real-time monitoring system will face tighter budgets and schedules—especially in a stage recently attacked with tokens and revenues both under pressure. This invisible squeeze may not appear in any financing news but is likely to manifest in the frequency and destructiveness of the next attack.

Purrlend's experience crystallizes this structural imbalance onto one ledger. It chose to deploy business on emerging networks like HyperEVM and MegaETH still in early ecological stages: the former incurred a loss of about $1,197,488.33, including about 449,683 USDC; the loss on the latter was at the hundreds of thousands level, estimated at about $324,549.49 by a single source, still pending validation. The new networks promote performance, low latency, and innovative architecture as their selling points, but behind these metrics, the coverage of security audits, on-chain monitoring infrastructure, and community emergency experience is significantly more limited compared to mature public chains—this is a reality repeatedly referenced in public information.

In this environment, the script of an attack often unfolds like this: protocols rushing to launch on high-performance new chains; the toolchains, monitoring alarms, and governance processes of new chains have not been fully streamlined; security audits can only focus on "ensuring core functionalities" within limited time and budget; and when genuine attackers arrive, the chain is complete with yet-to-be-fully understood boundary conditions, essentially handing complexity and vulnerability packaged to the opponent. Purrlend's loss of approximately $1.52 million, though not the largest in absolute market terms, represents a typical cost of "performance-first, defense-lagging."

On the other end, Drift stands at the extreme: after an attack on a scale of hundreds of millions of dollars (according to a single source), product manager Minh Don announced in Discord that the team will use a few weeks to optimize the codebase and plans to launch a post-fork exchange in May or June 2026, positioning "code reconstruction + protocol restart" as a disaster recovery route. Over the past few years, the DeFi sector has already gone through similar routes many times—forking, rewriting, governance restarts—whether or not these efforts succeed hinges on two core points: first, tighter security audits than before the incident, and second, sufficient liquidity and community support to restore trust with real money.

However, so far, no public information has emerged regarding awarded third-party security audits for Drift or Purrlend after this round of events, nor has a comprehensive compensation plan been disclosed. This means that when users weigh the decision of whether to "return," they have one hand holding unfulfilled technical promises and the other hand looking at still incomplete security and compensation details. And in the same time window, while AI companies use high valuation financing to bolster teams and expand R&D, DeFi protocols are filling holes left by breaches time and again with future revenue and governance rights.

Emerging networks like MegaETH and HyperEVM are the frontlines of this contradiction. They represent the imagination of “next-generation performance,” yet carry a capital volume far exceeding their own ecological maturity: funds priced today with future throughput, while protocols back their experiments with users' deposits. The systems of security governance, monitoring infrastructure, and community crisis plans are clearly lagging behind the speed of "TVL" and narrative expansion. As the AI sector continuously refreshes valuation ceilings and siphons off more developers and audit resources, the "performance-security" misalignment will only be further amplified.

Funds and talent are being pulled by another narrative, leaving behind protocols and new chains with increasing complexities and dwindling safety buffers. The next black swan may not necessarily fall on Drift or Purrlend; but it is certain that behind AI's high valuations, DeFi is continuously paying the cost for resource scarcity and safety discounts.

The next breaking point: Talent wars continue, cross-chain attacks persist

When you place DeepSeek's valuation of about $20 billion in this round, which has yet to announce financing details, alongside the on-chain safety incidents exposed by Purrlend and Drift, you will find that they are essentially telling the same story: the speed of expansion is exceeding the capacity of governance and risk control. The former is desperately squeezing into the "head AI company club," while the latter is racing against time on emerging networks and multi-chain deployments, with Purrlend facing a total loss of about $1,522,037.82 on HyperEVM and MegaETH, and Drift being forced into code restructuring and fork restarting after a loss exceeding $285 million. Capital, talent, and attackers are all migrating rapidly, while regulations and security practices inherently suffer from time lags and information asymmetries.

The coming months will likely see a constant stream of AI companies announcing high valuation financing on one side and new public chains and cross-chain protocols being persistently "tested" by hackers on the other. If DeepSeek successfully completes this round of financing, it will further align with high valuation trajectories like Moonshot AI; Moonshot AI completed three financing rounds in the first three months of this year, with its valuation quadrupling in a short time, while Zhizhu Technology and MiniMax have already listed on the Hong Kong Stock Exchange—this path's logic is clear: first enlarge valuation and story, then cash it out in the capital market. Yet, while valuations are accelerating, it becomes difficult to simultaneously upgrade internal governance, the sustainability of technical routes, and the clarity of profit models, especially when core talent can be lured away by larger checks at any time.

The on-chain world is playing out another isomorphic drama. Purrlend chose to deploy business on emerging networks like HyperEVM and MegaETH in their early ecological stages, gaining a faster growth curve and higher yield narrative, which also exposed it to the areas of limited security audit coverage and immature monitoring infrastructure, ultimately allowing attackers to find weak points. After the over $285 million attack, Drift's product manager Minh Don presented, on Discord, a restart window defined as "optimizing the codebase within a few weeks and launching a post-fork exchange in May or June 2026," attempting to regain trust through technical self-rescue and narrative reconstruction—but every instance of "rebirth" means making choices anew between accelerating iterations and raising the security budget. In the past few years, similar examples of DeFi relying on "coming back after incidents through forks and governance restarts" have been seen often, and success or failure highly depends on the quality of security audits and whether community liquidity will support funding—this will be no exception for Drift.

When both AI and crypto tracks enter such high-pressure states, the truly meaningful question is no longer who will become the next breaking point but who can proactively slow down their expansion pace a bit before the point of explosion, moving governance and security forward. From the reader's perspective at this moment, at least three dimensions deserve continued attention:

● First, watch for the "real technology" and "real profitability" of AI companies.
For players like DeepSeek, if they push for financing with a valuation of about $20 billion but repeatedly fail to disclose specific amounts and timelines, outside observers should return to the fundamentals: in which scenarios have their models and systems formed reusable technological barriers? How far are they from verifiable cash flow? In an environment where the talent war is intensifying and flows of top researchers are accelerating, companies that can only rely on "high valuations" and "star teams" to tell stories, yet fail to clearly articulate commercial paths, face increasing risks.

● Second, pay attention to the safety budgets and audit transparency of DeFi protocols.
After the Purrlend incident, any protocol deployed across multi-chain and multi-network that can only present yield curves yet cannot clarify how many rounds of high-quality audits they have completed, whether audit reports are public, or whether there is regular investment in monitoring systems and vulnerability rewards, is practically inviting attackers to "give it a try." Drift's choice to seek a second chance through code optimization and fork restarting reflects it paying for previous safety discounts—in assessing such "restarts," users should look beyond team reviews to see: this time, was the security budget truly placed at the forefront.

● Third, observe the self-upgrade of emerging networks in security governance.
Emerging networks like MegaETH and HyperEVM, still in their early ecological stages, tend to become targets for attackers faster than they can establish complete governance systems. Purrlend's damage on these two chains is merely an amplified signal: beneath performance and innovation, do these networks treat security as ongoing infrastructure development rather than a one-time audit before listing? Do they have clear emergency plans, public and transparent event disclosure mechanisms, and specialized governance for systemic weak links like cross-chain protocols? In the coming months, those who can actively provide these upgrade actions will be more likely to avoid the next round of tests.

The next breaking point may not happen with DeepSeek, Purrlend, or Drift; news headlines may simply read "Certain AI unicorn secures another high valuation financing" or "Certain cross-chain protocol hacked." What remains fundamentally similar is the underlying narrative: systems are being pushed to run increasingly fast while the hand controlling risks is always a step behind. For investors, developers, and ordinary participants caught in this dynamic, the most pragmatic protection is not to bet on which project will not run into trouble but to identify as early as possible those teams and networks willing to slow down their expansion pace a bit, moving governance and security forward one notch—because in a world where AI and crypto accelerate in parallel, those brave enough to "slow a bit" will paradoxically have a better chance to endure longer.

Join our community, let’s discuss and grow stronger together!
Official Telegram community: https://t.me/aicoincn
AiCoin Chinese Twitter: https://x.com/AiCoinzh
OKX benefit group: https://aicoin.com/link/chat?cid=l61eM4owQ
Binance benefit group: https://aicoin.com/link/chat?cid=ynr7d1P6Z

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。