Akash Moves to Solana: The AKT Migration Goes Live

On 12 November 2026, Akash Network opened a migration portal, and for the first time AKT began trading as a token on Solana. The move settles a debate that ran for more than a year over whether crypto’s best-known decentralized GPU marketplace should keep running its own Cosmos-based blockchain or move onto someone else’s. Governance gave its answer on 2 November, when Proposal 346 passed with about 97 percent support. The portal turns that vote into working plumbing.

And yet the token did not celebrate. AKT trades near $0.5698, down roughly 8 percent on the week, with a market value around $170.8 million. The whole decentralized-compute basket is in the red; Solana itself is off about 10 percent over seven days. A migration that stakers approved almost unanimously has landed in a market that is looking elsewhere. That gap, between a network doing something real and a token that struggles to price it, is the thread running through this entire story.

Here is what actually changed this week, why Akash walked away from the chain it built, what holders and stakers should do next, and whether moving to Solana does anything for the economics that have kept AKT more than 90 percent below its 2021 peak.

The migration portal is live

The headline is simple. As of 12 November, anyone holding AKT on the Cosmos chain can route it through an official portal and receive an equal amount of SPL AKT, the Solana token standard, at a one-to-one rate. The swap is one-way, and the portal is scheduled to stay open for two years. Akash’s own messaging to holders was blunt: there is no need to rush.

The new token carries six decimals and no freeze authority, and it is anchored to a single canonical mint address, 5d3fjPDdkYbtYcAMR9PC2TANaEyhzuyP8JxTEwqjVgBZ, published by the team. According to SolanaCompass, a Solana venue called Sunrise listed SPL AKT for trading within five minutes of the portal opening, describing it as the canonical AKT on Solana. Early liquidity was predictably thin: roughly 1.6 million SPL AKT had been minted by the evening of 12 November, backed by only about $244,000 in pool liquidity.

The crucial caveat is that this is the token phase only. The compute marketplace itself, the reverse auction that matches tenants with GPU providers, still runs on the existing Akash chain. Moving the token is the opening act; moving the business is a separate, later decision. For now, two versions of AKT exist side by side, and will for some time.

One detail worth underlining is the word canonical. In a migration like this, nothing stops a third party from issuing its own wrapped version of AKT on Solana, which is why Akash published a single official mint address and why the first listing venue went out of its way to call its market the canonical one. For holders, that means price discovery now happens in two places at once, the established Cosmos markets and the new Solana pools, and the two can drift apart while liquidity on the Solana side is still shallow. Confirming that you hold and trade the official mint, rather than a copy, is the whole game in these early weeks.

How the one-way swap works

Mechanically, the portal locks your AKT on the Cosmos chain and mints the same number of SPL AKT on Solana. Crypto.news reports that the process typically takes two to three minutes and carries a small network fee of about 0.0125 AKT. Because it is one-way, there is no sending SPL AKT back to Cosmos later; the lock is permanent, and the portal exists to move supply in a single direction over its two-year window.

Staked AKT is the wrinkle. If your tokens are bonded to a validator, you cannot migrate them until you unbond, and unbonding on the Akash chain takes 21 days. That matters for timing, and it is the reason Akash is telling stakers not to move yet (more on that below).

One-way portals are also a reminder to be careful. A cross-chain migration is exactly the kind of event that attracts fake-portal lookalikes and drainer sites, so the single most useful habit is to verify the canonical mint address before approving anything, and to treat any unexpected migration prompt as hostile until proven otherwise. The broader discipline here is the same one covered in our guide to crypto phishing defense: check the contract, slow down, and never let urgency do your thinking. A portal with a two-year window gives you all the time you need to get it right.

Why Akash walked away from the chain it built

To understand why a project would uproot its own token, it helps to know how Akash got here. Overclock Labs, founded by Greg Osuri and Adam Bozanich, launched Akash in 2018 as a permissionless cloud compute marketplace. Mainnet arrived in 2020 for CPU and storage, and the GPU marketplace that made Akash an AI-era name came in 2023 with the release it branded Supercloud. Throughout, Akash ran its own sovereign chain built on the Cosmos SDK.

The decision to leave started in late 2025, when the team concluded that maintaining a dedicated blockchain was a poor use of capital for a network whose actual job is renting out graphics cards. The accelerant came in 2026, when the license on a critical Cosmos software component was changed from a permissive open-source license to a source-available evaluation license that barred commercial use without a separate deal. Osuri called the change hostile on X, writing that it would prevent Akash from deploying the component in production or offering it as a service to third parties. For a team already questioning the cost of its own chain, that was the push.

The underlying argument is about capital efficiency. A proof-of-stake chain pays validators to lock up the native token and secure consensus. Akash was doing that for a network that settles only a modest number of transactions a day, which meant a lot of AKT was bonded purely to protect infrastructure, not to do work. Osuri has framed the whole pivot, as reported by crypto.news, around a long-standing goal of making compute a liquid resource. Shared security, borrowing another chain’s validators instead of paying for your own, is how he wants to get there.

Why Solana won the search

Akash says it evaluated seven options, with an EVM layer-2 as the runner-up, before choosing Solana. In the official writeup, Why We Selected Solana for Akash, the team lays out four reasons that are worth reading closely, because they double as a candid admission of what the old model cost.

The DePIN point is the strategic one. Akash is joining a neighborhood where decentralized physical infrastructure networks already live, and where Helium’s 2023 move off its own chain offers a template for how to do this without breaking everything. The liquidity point is the commercial one: a token that lives where the deepest stablecoin pools and lending markets sit is easier to price, borrow against, and integrate.

Akash is also honest about the trade. The blog explicitly lists what the network gives up by leaving: sovereign consensus, validator income, native inter-blockchain communication, and the story of owning its own chain. Those are not trivial losses, and one of them, validator income, sits at the center of the question every AKT holder should be asking.

The choice was not a foregone conclusion. Akash has said an Ethereum-style layer-2 was the runner-up, and it laid out specific conditions that would have flipped the decision the other way, including large exchanges refusing to support Solana’s newer token standard or the chain’s performance upgrades proving unstable in testing. None of those conditions triggered, so Solana held. That kind of pre-committed decision rule is unusually transparent for a crypto governance process, and it matters because the same discipline is meant to govern the riskier steps still to come, the audits, the parameter settings, and the eventual marketplace cutover.

What stakers should do right now

If you hold unstaked AKT, nothing forces your hand. You can migrate today, migrate next year, or sit tight; the portal stays open for two years and the swap rate is fixed at one-to-one. There is no deadline pressure and no bonus for going first.

If your AKT is staked, the advice from Akash is different: stay put for now. Migrating staked tokens means starting the 21-day unbonding clock first, and during that time you earn nothing. More to the point, staking rewards continue on the Akash chain right up until the marketplace cutover, which has not happened yet. So a staker who rushes to Solana today gives up ongoing Cosmos rewards in exchange for a token that, at this moment, can only be transferred and traded.

That is the part many holders will miss in the excitement. The SPL version of AKT launched with no staking and no governance rights attached. It is, for now, a tradable asset and nothing more, and Akash has said the details of staking SPL AKT on Solana are still to come. Until that model ships, migrating early is a bet on liquidity and convenience, not on yield.

There is also a practical reason not to unbond in a hurry. The 21-day exit queue means a staker who starts migrating today is out of the rewards stream for three weeks and holding an idle position, all to arrive at a token that cannot yet be staked on the other side. Unless you need the liquidity now, or you specifically want your AKT sitting in a Solana wallet for a decentralized-finance use, the path of least regret is to let the token phase settle, watch for the SPL staking design, and move once there is something productive to move into.

The staking sink is going away

This is the analytical heart of the migration, and it cuts against the celebratory framing. On a proof-of-stake chain, bonding is a demand sink: tokens get locked up and taken out of circulation because locking them earns rewards and secures the network. Shared security removes that. Once Akash settles on Solana, it no longer needs roughly 100 validators to bond AKT to protect a chain, because Solana’s validators already do the securing. Capital that was previously parked in AKT for the sole purpose of consensus is freed. Efficient, yes. But it also deletes one of the few structural reasons to hold and lock the token.

This is not an abstract concern. A large share of what makes a staking token worth holding is precisely the machinery that pays people to lock it up, the subject we dug into in our piece on validator economics. Take the lock-up incentive away and you are relying on pure transactional demand, which for AKT means people burning it to pay for compute. Whether that demand is large enough to matter is the open question the rest of this article keeps returning to.

It is worth noting that AKT staking was not a generous trade to begin with. New issuance is capped in the single digits annually, and about half of it flows to a community pool rather than to stakers, which has kept real staking yields negative once inflation is counted. As with liquid staking tokens elsewhere, the headline rate and the yield you actually keep are very different numbers, a gap we unpacked in our comparison of Lido, Rocket Pool and Frax. So the sink that is disappearing was already leaky. Still, a leaky demand sink is more than none, and until the Solana staking model is defined, the one on-chain reason to hold-and-lock AKT is simply paused.

What is supposed to replace it is transactional demand. In the Solana design, the reason to want AKT is that tenants spend it, or stablecoins, to rent compute, and the burn-mint system routes that spending through the token. The bet is that a cheaper, more liquid home will grow real usage faster than the old staking subsidy ever locked up supply. It is a coherent bet, and it is the same one every fee-driven token makes. It only pays off if the marketplace underneath generates enough throughput that the token is genuinely needed, rather than merely passed through and sold.

Four more votes before the job is done

Proposal 346 was decisive but narrow. It authorized the token phase, not the full move. According to NullTX, which tracked the final tally at 59,244,540 AKT in favor against under a million opposed, with turnout near 69 percent of bonded stake, four more on-chain votes still stand between today and a finished migration.

1. A software-upgrade proposal that adds the sink module to Akash, with deposits initially disabled.
2. A parameter vote that sets the reporters, caps and challenge period, and turns deposits on.
3. A marketplace cutover proposal that decides when the compute business itself moves to Solana.
4. A sunset upgrade, filed 21 days before the old chain halts, that formally winds the Cosmos chain down.

The timeline matters as much as the sequence. The proposal mandates two independent audits of the token and portal code, with reports published within 30 days. The marketplace will keep running on the Cosmos chain for about 90 days after the cutover vote passes, so existing deployments do not break mid-stream, and only then does the original chain halt. Akash itself describes the whole arc as roughly a two-year process. The right way to read this week’s news, then, is as the first domino, not the last. AKT moves first; the marketplace moves later; and anyone saying Akash is finished moving to Solana is getting ahead of the governance calendar.

Real compute, a token that struggles to capture it

Here is the uncomfortable fact of the week: AKT fell during its own Solana debut. Part of that is sector beta, with the entire decentralized-compute group and Solana itself trading lower. But part of it is a problem Akash has not solved, and moving chains does not fix it. The token captures the marketplace’s value only weakly.

The mechanism is the burn-mint equilibrium that Akash activated in March 2026. A tenant burns AKT to buy a settlement credit pegged to US dollars, and the provider is paid AKT at the settlement price. The intent is elegant: usage should translate into AKT being destroyed. In practice, the net on-chain destruction is close to zero unless AKT appreciates between the moment it is burned and the moment it is settled, because the same economic value is minted back out to pay providers. It is, in effect, a burn that barely burns.

The critique has teeth because Akash has never published how much AKT the burn side actually removes on net. Zoha Imdad Ali, writing for TECHi, made the point directly: Akash runs real AI compute, but usage and token value are separate things, and the network does not disclose the net figure that would let anyone check whether the token captures the value. Until that number is published, the burn narrative is a story, not a verified line of defense for the price.

All figures above are from CoinGecko on 13 November 2026. A token trading 93 percent below its cycle peak is not unusual in this corner of the market, but it is a reminder that none of the technical progress of the past year, confidential compute, a usage rebound, and now a chain migration, has re-rated AKT in any durable way.

The business is quietly rebounding

The gloom around the token obscures a genuinely better operating picture. After a year of contraction, Akash’s numbers turned up in the third quarter. Per Messari’s State of Akash Q3 2026 summary, shared via the firm’s official account, protocol revenue rose about 73 percent quarter over quarter to roughly $304,000, GPU capacity grew 8.5 percent to 420 units, and GPU utilization climbed 42 percent quarter over quarter. Messari credited rebounding deployment activity, steady inference demand, and a round of Console, API and billing upgrades that lowered the barrier for developers to experiment.

To see why that is a rebound, you have to look at where Akash was. The first quarter of 2026 was the trough, and it was ugly.

Those Q1 figures come from Messari’s Q1 report, which also logged 43,540 new leases, up 27 percent, even as revenue fell, and full-year 2025 revenue of $3.15 million, up 128 percent. The rebound is real, but keep the absolute scale in view: a quarter near $304,000 is a strong bounce off a low base, not evidence of a business suddenly minting money.

There is also a transparency gap worth flagging. Akash’s own materials have at times pointed to a cumulative compute milestone far larger than the on-chain lease revenue Messari counts, spanning its managed products as well as raw marketplace leases. The company has never published a reconciliation between the two numbers. That is most likely a difference of scope and timeframe rather than anything sinister, but a project asking investors to value a token on usage ought to show its work.

Real demand, thin supply

The rebound is being driven less by the old picture of researchers renting bare GPUs and more by inference. Over the past year Akash has leaned into a stack of managed products, AkashML, Akash Chat and Akash Agents, that serve open models such as Llama and DeepSeek directly, so a developer can hit an endpoint instead of writing a deployment manifest. That is the demand engine Messari means when it says inference held steady.

It helps to be honest about the size of that engine. A Blockworks survey, flagged via KuCoin, found that seven crypto-native inference providers together process around 211 billion tokens a day, which is roughly 1 percent of what a mainstream router like OpenRouter handles. Within that small group, Akash’s AkashML did not even disclose a figure and is not the largest provider. The demand is real and the report noted it still exceeds supply, but Akash is a minor player inside a niche, not the center of AI compute.

The deeper constraint is supply. Akash’s active provider count hit a record low this year, and its reverse auction, where providers bid against each other and tenants pick, pushes rental prices toward the floor. That is wonderful for buyers and brutal for a provider trying to pay off a card that costs tens of thousands of dollars while it sits at roughly a third utilization. The economics simply do not close for a small operator at those prices.

Akash’s answer is a pincer. Top-down, it is standing up StarCluster, a protocol-owned GPU mesh financed through a regulated US investment instrument and built around several thousand Nvidia Blackwell GPUs run by vetted operators. Bottom-up, Homenode lets owners of consumer cards like the RTX 4090 and 5090 become providers without running Kubernetes. A separate incentive program would pay providers in AKT to show up, which adds supply but also dilutes holders. And a trust note runs underneath all of it: Akash shipped confidential computing this year so a tenant can hide data from an anonymous host, but confidential is not the same as verifiable. A provider can keep your data secret and still run the wrong model, and the trustless alternatives, the proof systems we examined in our look at opML and verifiable AI, remain far too heavy for production inference today.

How Akash stacks up against the field

Akash is not alone in the decentralized GPU trade, and the migration to Solana puts it in the same settlement layer as several rivals. Render pairs rendering with inference and already lives on Solana; io.net focuses on clustering and orchestration; Nosana is a Solana-native inference play; Phala leans on trusted hardware for confidential workloads. Bittensor is sometimes lumped in but is a different model built on subnet rewards, so it sits outside this comparison.

Token prices above are from CoinGecko on 13 November 2026, and the whole group traded lower on the week, which is the clearest sign that this is sector sentiment, not an Akash-specific verdict. The more sobering comparison is with the centralized benchmark. CoreWeave, which rents the same kind of hardware without a token, reported third-quarter 2026 revenue of $3.66 billion, up 168 percent year over year, with a revenue backlog of $132.5 billion. The entire decentralized GPU basket, Akash included, generates an annualized revenue that would not register against a single CoreWeave quarter. CoreWeave’s chief executive, Michael Intrator, has repeatedly told investors that the real bottleneck in AI is no longer whether companies want to deploy but how fast reliable capacity can be delivered. Decentralized networks are betting they can supply some of that capacity more cheaply; the scale gap shows how far they have to travel.

The macro backdrop is the one thing working in every GPU network’s favor. Demand for AI compute has run well ahead of the power and data-center capacity available to serve it, which is why Bitcoin miners have spent the year converting megawatts into AI hosting and why centralized providers keep signing multi-year capacity deals. In that environment, a network that can aggregate idle and underused GPUs has a real reason to exist. The question for Akash is never whether the demand is there; it is whether a permissionless marketplace can marshal reliable supply at the moment a serious customer needs it, which is exactly where its thin provider base still bites.

Where the SEC fits

For a US holder, the migration is mostly a change of form, not substance. AKT reads as an infrastructural utility token: you spend it, or stablecoins, to rent compute, and moving it from a Cosmos chain to an SPL mint on Solana does not change what the token does. The Securities and Exchange Commission has taken a lighter-touch posture toward this kind of utility asset in 2026, and the fact that tenants can pay for compute in USDC keeps many users off the token entirely.

The clearer securities question sits elsewhere. Akash chose to finance its StarCluster hardware through an instrument it structured as a regulated US investment product, which is explicitly a security and is marketed as one. That is a cleaner line than many crypto projects draw: the speculative, yield-bearing instrument is registered, while the network token is positioned as a utility. Holders should still watch how the new SPL staking model is designed, because the moment a token starts paying holders a return for locking it up, the question of how that return is generated, and whether it looks like an investment contract, gets sharper. For now, the migration itself does not create a new US registration event.

There is a tax dimension too, and it is unsettled. Moving a token one-to-one from one chain to another can read as a non-event or as a disposal depending on how a given jurisdiction treats it, and US holders should not assume a 1:1 bridge is automatically tax-free. The conservative approach is to keep records of the migration date, the amounts, and the cost basis of the original AKT, and to treat the SPL tokens as a continuation rather than a fresh purchase unless a tax professional advises otherwise. None of that is Akash-specific, but a two-year window means a lot of holders will cross this line at different times, with very different prices on the clock.

What to watch next

The next few months will decide whether this migration is a genuine reset or a change of address. A short checklist is enough to track it.

• The two mandated audits of the token and portal code, and their reports, due within 30 days.
• The four remaining governance votes, especially the marketplace cutover that moves the actual business.
• The Solana staking model for SPL AKT, the missing demand sink, and how it is designed.
• Liquidity depth on Solana; early pools are thin, and a one-way portal is also a bridge, with its own risk surface.
• Fourth-quarter Messari data, to see whether the Q3 rebound holds or fades.

That bridge point deserves a flag of its own. A one-way migration portal concentrates a lot of value behind a small amount of code, and 2026 has been a hard year for anything that moves tokens between chains, as we covered in our piece on the two-speed split in bridge security. The two mandated audits exist precisely because this is the riskiest surface in the whole plan. For holders, the honest summary is that Solana gives AKT a bigger, cheaper, more liquid venue, and a neighborhood of DePIN peers to grow alongside. What it does not do, on its own, is answer whether the token can finally capture the value its marketplace produces. That remains the question, and the migration just gives it a new place to be asked.

Frequently Asked Questions

Is AKT now a Solana token?

Partly. Since 12 November 2026, holders can move AKT to an SPL version on Solana through a one-way portal, and it already trades there. The original Cosmos-based AKT still exists and still secures the marketplace, and the portal stays open for two years, so both forms coexist during the transition.

Does migrated SPL AKT earn staking rewards?

Not at launch. The Solana version supports transfers and trading only, with no staking or governance rights for now. Akash has said a staking model for Solana will arrive before the marketplace cutover, and it advises current stakers to keep their tokens bonded on the Cosmos chain, where rewards continue in the meantime.

Why did Akash leave Cosmos for Solana?

Akash concluded that paying roughly 100 validators to bond AKT just to secure its own low-traffic chain was capital inefficient, and a 2026 change to a Cosmos software license pushed it to act sooner. It selected Solana for cheap fees, deep liquidity, and the cluster of other DePIN networks already settling there.

Will moving to Solana make AKT go up?

There is no guarantee. AKT fell during the week of its Solana debut, and the migration removes the staking demand that locked tokens on the old chain. A more liquid home can help, but the token still has to capture the value its marketplace produces, and that remains the open question.

Is the Akash marketplace already running on Solana?

No. Only the token has started to migrate. The compute marketplace stays on the existing chain until a separate governance vote approves the cutover, after which older workloads keep running for about 90 days before the Cosmos chain halts. The full transition is expected to take roughly two years.

Marcus Okafor is a senior markets writer at HOGE Wire, covering decentralized infrastructure and the economics of crypto networks.