Akash Activates Burn-Mint Model to Link Compute Demand With AKT Value

Decentralized cloud marketplace Akash has entered a pivotal new phase after switching on a token-economic framework designed to tie real network usage directly to token value—an attempt to turn compute demand into sustained 'AKT buy pressure' and potential 'deflationary supply' over time.

In a May 15 report, Messari Research analyst Eric Manoukian detailed how Akash activated a 'Burn-Mint Equilibrium' (BME) model via its Mainnet 17 upgrade on March 23. The change links all on-chain compute spending to market purchases of Akash (AKT) followed by token burns—marking the network’s first explicit deflation mechanism embedded into core economic flows. AKT rose 41.6% over the first quarter to end at $0.50, with much of the rally clustering around the BME activation window, a market pattern Messari said may indicate investors are beginning to reprice AKT as a function of real compute usage rather than purely speculative demand.

Akash is a Cosmos SDK-based layer-1 network that aims to serve as an alternative to centralized hyperscalers by matching providers with unused server capacity to tenants that need compute. The platform supports containerized workloads ranging from AI tasks and game servers to blockchain nodes and websites, and has increasingly positioned itself as an 'open compute layer' for decentralized AI applications and agent deployments amid expanding GPU demand.

The core issue BME seeks to address is what Messari described as a structural disconnect under the prior AEP23 payment framework. Tenants paid for compute using axlUSDC, giving providers stable dollar-denominated settlement—an advantage for enterprise adoption—but the arrangement did little to create direct demand for the network’s native token. In effect, usage could grow without meaningfully affecting AKT demand.

Under BME, compute spend is first converted into AKT via market buys, then burned. In parallel, the system mints ACT, a non-transferable, dollar-denominated payment token held during the lease term. Tenants and providers continue to transact based on dollar pricing, but internally each unit of compute consumption is routed through AKT market demand. Providers ultimately redeem ACT into AKT at settlement based on prevailing market prices. If AKT appreciates between tenant funding and provider settlement, the mechanism can result in net burns that reduce circulating supply—meaning higher network usage can translate into stronger 'supply-side compression'. Messari characterized this as the largest economic design shift in Akash’s history.

BME was enabled through Mainnet 17 (v2.0.0), which implemented four proposals at once—AEP76, AEP78, AEP80, and AEP81. AEP76 defined the BME structure, while AEP78 activated CosmWasm smart contracts to make the logic auditable and upgradeable. AEP80 introduced a native oracle module to support price aggregation and time-weighted average price (TWAP) calculations, and AEP81 integrated AKTUSD pricing from Pyth Network through Wormhole’s verification pathway. The multipronged pricing design is intended to mitigate manipulation risk from reliance on a single source.

Governance around the shift was decisive. An incentivized testnet began on February 17 with roughly 250 participants and more than $10,000 in rewards across over 10 testing categories. The Mainnet 17 upgrade proposal, No. 318, passed with 99.7% approval in a vote held March 6–13. The upgrade executed at block 26,063,777 on March 23 and removed the legacy x/take module. By March 31, the network had burned 53,520 AKT under the new framework.

Price action followed quickly. AKT started the quarter near $0.37 on Jan. 1, slid to about $0.28 on Feb. 5, then surged into late March—reaching around $0.60 on March 21 before closing the quarter at $0.50. Circulating market capitalization rose 30.2% quarter-on-quarter to roughly $130.7 million from $104 million. Messari emphasized that the rally’s timing—concentrated around March 23—suggests traders may have begun pricing in a new value-capture route tied to compute consumption.

However, usage and revenue metrics did not uniformly confirm a broad-based rebound. New leases rose to 43,540 in Q1, up 27.1% from the prior quarter and marking a third consecutive quarterly increase. Yet lease revenue fell sharply to $253,250 from $460,510, a 45% decline. Average active leases dipped 4.4% to 583 from 610, suggesting many new leases may have been smaller, lower-cost workloads rather than higher-value deployments.

The squeeze was also visible in network fees. Lease revenue accounted for 98% of total network fees of $257,580, but total network fees fell 44% quarter-on-quarter. For Messari, this underscores that Akash remains heavily dependent on leasing activity, with transaction fees and other revenue sources not yet scaling meaningfully. In that context, BME may have fixed the 'value capture plumbing', but it still requires higher absolute compute consumption to generate substantial and sustained token-economic impact.

On the supply side, the network also contracted. Average active providers fell 8.4% to 58 from 63—the lowest level in recent quarters. Capacity and usage declined across GPU, CPU, storage, and RAM. Average GPU usage dropped 57.4% to 84 units and available GPU capacity fell 57.5% to 334, though utilization held steady at 33.7%. CPU usage declined 21.1% to 2,420 vCPUs while capacity fell 46.5% to 11,690 vCPUs, lifting CPU utilization from 17.7% to 26.1%—a sign providers reduced idle resources faster than demand fell.

Storage utilization remained particularly low: average usage fell to 23.5 TB and capacity to 646.3 TB, leaving utilization around 3.6%. RAM usage declined to 5.4 TB with capacity at 66.5 TB, keeping utilization near 8.1%. Messari attributed the imbalance to Akash’s current workload mix, which is more GPU/CPU-centric due to AI inference and related tasks, with comparatively less storage- and memory-intensive demand on the network.

Still, Messari pointed to late-quarter product launches as potential catalysts for renewed compute consumption. On March 26, the network launched Akash Agents, a platform that enables one-click deployment of AI agents such as OpenClaw and Hermes, lowering friction for developers and teams experimenting with autonomous AI workflows. The report said daily lease counts accelerated after the release, suggesting the product could translate ongoing AI agent interest into measurable network usage if adoption persists.

Akash also began early access for Akash Homenode on Feb. 25, targeting a broader supplier base by allowing individuals to connect consumer and prosumer GPUs—such as RTX 4090, RTX 5090, and RTX 6000 Ada-class hardware—from home. Where Akash’s provider network has historically skewed toward data-center operators, Homenode is positioned as an experiment in democratizing supply, reducing dependence on centralized infrastructure, and potentially enabling more geographically distributed AI inference capacity.

Protocol work continued in parallel. Mainnet 16 went live March 4 with CometBFT Tachyon security fixes, store migration work, and a feature to record on-chain reasons for lease termination—preparatory changes ahead of BME’s rollout. The network also approved seven notable governance proposals during the quarter, focusing on GPU capacity support, market-making reinforcement, BME engineering funding, and groundwork for the Homenode MVP.

Community activity leaned into the AI infrastructure narrative through events such as AI Agent Build Night, Continual Learning Hackathon, Open Agents Hackathon, and participation at the Penn Blockchain Conference, an approach that appears aimed at expanding long-term developer and startup engagement rather than driving short-term token momentum.

Overall, Messari’s assessment frames Q1 2026 as a structural turning point: Akash has re-engineered its token economics so that real compute demand can more directly influence Akash (AKT) through market buys and burns. Yet the report also highlights clear execution challenges—declining revenue, fewer active providers, and shrinking capacity suggest the network must still prove it can scale high-value workloads. In that sense, AKT’s longer-term trajectory may depend less on the novelty of BME itself than on whether products like Akash Agents and Akash Homenode can convert the broader AI compute boom into sustained on-chain consumption.