From the zero to tens of thousands of GPU computing power, to encountering dozens of DDoS attacks, and then to secret talks with government officials from various countries about AI sovereignty—this is a true story about belief, decentralization, and human gifts.
Source: DeAI Nation's "State of DeAI 2026" report | Compiled by: Gonka.ai
David Liberman is the co-founder of Gonka, a decentralized AI inference and training network. Launched in August 2025, the network quickly amassed over 10,000 GPUs (equivalent to an NVIDIA H100) within months. This article is based on an interview with David in the DeAI Nation's "State of DeAI 2026" report, covering his in-depth analysis of the "AI version of Bitcoin" argument, the controversies surrounding the boundaries of decentralization, the numerous attacks Gonka has faced, and transcripts of his discussions with various governments on AI sovereignty.
I. Everyone wants to become the Bitcoin of the AI world.
When observing the decentralized AI ecosystem, one phenomenon stands out: almost every project and every blockchain is trying to position itself as the new Bitcoin in the world of artificial intelligence. Is this simply a matter of inertia within the crypto industry, or is there some deeper structural reason behind it?
David offered his assessment: it's a combination of two motivations, just with different emphases in different projects.
From a comparative perspective, this kind of analogy isn't unique to the crypto world; it's prevalent throughout the entire tech industry and even the entire startup ecosystem. When a disruptive new thing emerges in a field, pioneers must prove the feasibility of the underlying logic from scratch, while later entrants can stand on the shoulders of giants, using existing success stories to endorse themselves. Just as Silicon Valley investors require entrepreneurs to introduce themselves at the beginning of their fundraising presentations like this: 'We are the Airbnb for dog owners'—this short sentence saves the tedious effort of repeatedly demonstrating the feasibility of the platform economy model.
Bitcoin is neither the first decentralized project nor the first open-source project in history; in fact, BitTorrent was already a prime example of a decentralized network. What Bitcoin truly proves is that the incentive model based on tokenomics can function independently in the real world. The value of this proof allows all subsequent projects built on tokenomics to legitimately skip this crucial step of the argument.
"We use Bitcoin as an analogy, partly to avoid the hassle of re-proving the viability of token economics. There are still some skeptics who believe Bitcoin will eventually go to zero, although such people are becoming increasingly rare." — David Liberman
However, for Gonka, this analogy carries a deeper meaning. While most crypto projects have shifted to Proof-of-Stake (PoS) mechanisms, Gonka adheres to Proof-of-Work (PoW) and builds its computing infrastructure around it. David explicitly stated: Gonka is following the path of Bitcoin, not the path of modern Ethereum. Ethereum initially also adopted PoW, which spurred the development of mining infrastructure, but later shifted to PoS, gradually distancing itself from this infrastructure incentive system.
His assessment is that PoW can create stronger infrastructure incentives. Of course, it's also understandable that other projects choose to use Bitcoin as an analogy to describe themselves—the key point is that no one is claiming they can reach Bitcoin's market capitalization, but rather saying that the underlying propositions validated by Bitcoin also apply to us; the only new variable is AI.
II. How does Silicon Valley view decentralized AI?
When the concept of "decentralized AI" entered Silicon Valley, the reaction it evoked was far more complex than outsiders imagined—it included enthusiastic endorsement from crypto investors, profound reflection from the AI security research community, and silent observation from practitioners at large model companies.
David mentioned two representative voices: Chris Dixon, a partner at a16z, has long been a vocal supporter of decentralized AI and has made investments in the field; Shaun Maguire, a partner at Sequoia Capital, wrote that crypto and AI are a natural pair. Although some believe that Dixon's stance stems from his crypto background, these voices still constitute a positive testament to decentralized AI within Silicon Valley.
More noteworthy is the quiet shift within the AI security research community. David points out that almost all the foundational scientists of modern AI originated from the AI security research community. The birth of OpenAI itself stemmed from concerns about Google's monopoly on AI, and was intended as a counterbalancing alternative—but this original intention quietly crumbled as OpenAI itself gradually approached a monopoly.
"The AI security community once opposed decentralization, unwilling to release AI capabilities to ordinary people. But as computing power has become highly concentrated in the hands of a few giants, this community has begun to realize that without sufficient computing power, no AI security research can proceed. As a result, their attitude towards decentralization is undergoing a fundamental shift."
Meanwhile, among the broader developer community, the appeal of decentralized AI is becoming increasingly clearly linked to cost. David observes that when a project is just starting out and has VC funding, using centralized inference services is not costly; however, as the scale increases, the bills become a stark reality. He gives a vivid example: many developers integrated their AI agents with Claude Opus, only to find the next morning that the agents had been running all night, resulting in alarming token consumption, prompting them to urgently seek alternatives.
The changes in OpenRouter's data confirm this trend: two months ago, almost all of the top-ranked models on the platform were closed-source models; now, the proportion of open-source models has increased significantly. David's judgment is: "Every financial crisis pushes more people to Bitcoin, and the large-scale adoption of decentralized AI will unfold in the same way—wave after wave, each wave retaining more users. The first wave will be driven by price."
III. Where are the boundaries of decentralization?
While the entire industry is chanting the slogan of "decentralization," the term itself is quietly losing its precision. David admits that the concept has been diluted to varying degrees—partly because true decentralization is extremely difficult to achieve at the engineering level, and partly because some projects, under the guise of "incremental decentralization," actually maintain long-term control over core power.
He understands the trade-offs made by projects: "Whenever you claim to be completely decentralized, you'll encounter obstacles at every stage. Some projects say, 'We're not decentralized here, we're only decentralized there,' especially in AI infrastructure, where many early-stage projects have made excessive compromises. Personally, I believe that excessive compromises can sometimes damage the credibility of the decentralized concept itself."
Gonka's choice on this point is particularly clear: from the beginning, the team chose not to retain control for themselves, but to hand over governance to the community. This drew a lot of criticism from the outside world, but David always insisted: "Why must everyone trust a centralized authority? Decentralization is what truly attracts trust." The cost is real—every change must be discussed with everyone.
In David's view, there is a general, albeit imprecise, rule in this industry: projects with higher levels of decentralization tend to carry greater value. The market capitalization of Bitcoin and Ethereum has long exceeded that of XRP and even Solana; conversely, projects found to have their founders and foundations effectively controlling the entire ecosystem often lose a significant portion of their market capitalization as a result.
"Decentralization is not a marketing label, but a mechanism for building trust over the long term. Filters around power structures are real in this industry, although they don't always work in a timely manner."
He also explicitly expressed his respect for Prime Intellect, considering it an outstanding team that dared to confront the core challenge of decentralized training. However, he also pointed out that there is still no clear answer regarding the business model of decentralized training—because the continuous emergence of more capable free and open-source models makes competition in the training market increasingly difficult. Gonka's ultimate decision to focus on inference is based on a sober assessment of business realities: inference generates continuous demand, fosters real-world infrastructure, and is precisely the direction that capital is truly willing to flow into.
IV. Attack, Collapse, and Resilience: Gonka's Test of Life and Death
Since its launch in August 2025, Gonka has undergone far more intense stress testing than expected.
David admitted that Gonka suffered not a single DDoS attack, but dozens. The attacks began in the first month after the system went live, initially on a smaller scale and with simpler methods, but from late December 2025 to January 2026, the scale and complexity of the attacks significantly increased. The attackers continuously searched for every possible vulnerability, constantly testing the limits of the system.
This exposed the drawbacks of Gonka's highly decentralized design: in a centralized system, attacks can be coordinated and responded to directly by the core team; however, in a decentralized network, each miner must ensure the security of their own infrastructure. The network includes both seasoned crypto miners and many new participants attracted by the decentralized AI concept—the latter lacking experience and tools to combat cyberattacks. This made community-level security education a top priority.
At the peak of the attack, several nodes were still taken offline each day due to the attack. But a more serious problem came from the design of Gonka's original incentive mechanism: when a miner was unable to prove availability due to an attack, its daily reward would be forfeited and redistributed to the remaining miners—meaning that defeating 30% of the miners could increase one's own earnings by 30%. Attacks became profitable.
"We have personally experienced a paradox: decentralization makes us more vulnerable to attacks, but it also makes our defenses stronger through community participation."
The community subsequently voted to modify this mechanism, so that attacking others would no longer directly bring economic benefits. Attacks didn't disappear, but the underlying motivation was significantly reduced. David admitted that he now understands why some projects choose to centralize their APIs—a distributed, publicly accessible API node is far more difficult to protect than a centralized architecture. However, Gonka's stance remains unchanged: APIs should remain open and decentralized, as this is at the core of the entire project's philosophy.
Meanwhile, the sluggish macro crypto market has also put pressure on the industry. Bittensor's GPU count has declined, and Gonka's peak GPU count has also decreased. But David characterizes this period as a "breathing room": "If Bitcoin were at $120,000 today, the number and scale of attacks would likely be several times higher. Now is the best time to strengthen defenses before the next bull market arrives, while the market is calm."
Even after all this, the Gonka network still operates approximately $200 million worth of hardware assets online, with a significantly larger number of GPUs than other similar projects. David sees this as a concrete manifestation of the community's commitment.
V. Governments around the world discuss AI sovereignty: Computing power is power
Another parallel storyline in Gonka's development is equally compelling: David and Daniil frequently met with government officials and executives from large corporations to discuss the potential of decentralized AI at the national strategic level. These conversations reveal a grander vision that transcends mere business logic.
David observed that governments' interest in decentralized AI ultimately stems from three levels of motivation.
Motivation 1: Computing power sovereignty
Currently, government services in many countries are heavily reliant on AI, but the computing power behind it is controlled by external service providers. This dependence brings not only cost issues but also strategic risks: once external suppliers control access, pricing power, or infrastructure, they could use this as leverage to restrict or even cut off critical services. This structural vulnerability is the issue that government officials around the world are most concerned about.
Motivation 2: Development of the local industry
Governments around the world want their data center industries to truly take root locally, rather than simply becoming "cloud access points" for foreign companies. They expect local employment, local capital accumulation, and long-term technological capacity building—not handing over data and profits to a few hyperscale cloud service providers.
Motivation 3: Participation in the chip industry chain
Some countries are already looking further upstream: not only to operate data centers, but also to participate in semiconductor manufacturing. This is not wishful thinking, because the entry point is not the most advanced 3-nanometer process, but more mature process nodes such as 16-nanometer—which is realistically feasible for more countries.
At the intersection of these three motivations, the narrative of decentralized AI networks begins to demonstrate its unique persuasiveness.
"What we showed them was not just sovereignty, but a viable economic model. If a country participates in a decentralized computing network, it can build a data center with 20,000 GPUs and generate endogenous demand from the global market—instead of hoping that Microsoft or some hyperscale service provider will be willing to rent your computing power at a reasonable price."
David used Bitcoin as an analogy: Bitcoin achieved natural growth in computing power globally without any single country holding a structural advantage. Token economics creates distributed economic incentives, allowing countries to choose to participate independently without being dependent on a centralized ecosystem leader. He believes the same logic can be applied to the global distribution of AI computing power.
Of course, there are also practical complexities: local infrastructure often struggles to operate at full capacity 24/7, and idle capacity remains a persistent economic challenge. David's solution is a hybrid "local + distributed" model: while the local cluster handles the basic load, idle computing power is connected to a global decentralized network, turning idle resources into continuous revenue; during peak periods, additional computing power is drawn from the network to handle sudden demands. He cited the logic behind the creation of Amazon Web Services—it was the huge elastic computing power demand of e-commerce platforms during peak holiday periods that gave rise to the cloud computing business model, and today's AI computing power scheduling faces the same structural problems.
VI. The Other Side of Training: A Gift for Humanity
As a future vision, Gonka proposed allocating 20% of its inference revenue to decentralized model training. David expressed genuine anticipation for this, but also acknowledged the challenges involved.
He stated bluntly that decentralized training remains an unsolved engineering problem, and its commercial viability is almost unanswered. The reason is simple: the open-source community is constantly producing more powerful and completely free foundational models, which has virtually killed the market space for independent training. Any project attempting to commercialize decentralized training will find it difficult to compete with free open-source alternatives—unless your goal is to become a cutting-edge AI lab.
Gonka chose a different path: focus on inference first, building up the infrastructure and token economy to achieve true economies of scale, and then using a portion of the network's capabilities for training. The logic behind this path is: first, there must be computing power scale, then training becomes possible, rather than the other way around.
"Training may not be our growth engine, but it can be a gift we give to humanity. Why not? No one loses anything from it, and we have the opportunity to give the world something truly valuable."
David frankly admits that many prerequisites led to this point: physical engineering challenges, collective consensus within the community, and the continued growth of the Gonka network as a whole. He knows clearly that this won't happen anytime soon. But he also points out that any breakthroughs achieved by the teams that have poured tens of millions of dollars into this direction, working day and night, will ultimately belong to all of humanity—because replicating an achievement is usually far easier than achieving it for the first time. He has great respect for these teams and has positioned Gonka's primary mission as building a decentralized computing infrastructure capable of truly competing with top-tier cutting-edge labs and hyperscale service providers.
Conclusion
David Liberman's story is about entrepreneurs navigating treacherous waters—dealing with cyberattacks, proving the value of decentralized AI to hesitant government officials, and maintaining community faith amidst an uncertain crypto market cycle.
However, behind all of this, a clear theme runs throughout: decentralization is not a marketing slogan, but a philosophy of building infrastructure based on long-term trust. Gonka chose the most difficult path, and that's why it has come this far.
This experiment in decentralized AI is far from over. But as David said, the price paid by every pioneer becomes the starting point for those who follow. And those who persevere through the most difficult times will ultimately see the significance of their work in the next wave.
