22 posts tagged with "DePIN"

Every chip NVIDIA can make for the next two years is already spoken for. At GTC 2026 on March 16, Jensen Huang unveiled Vera Rubin — a 336-billion-transistor AI platform built on TSMC's 3nm process — while simultaneously confirming what the industry already feared: HBM4 memory is completely sold out through 2026, and GPU lead times now stretch 36 to 52 weeks. For the $19 billion DePIN sector, this supply crisis isn't a problem. It's the opportunity of a decade.

The Vera Rubin Architecture: A New Scale of AI Compute​

Named after the astronomer who proved the existence of dark matter, Vera Rubin represents NVIDIA's most ambitious platform leap since Blackwell. The numbers are staggering:

• 336 billion transistors on TSMC's N3P node — nearly double Blackwell's density
• 22 TB/s memory bandwidth via next-generation HBM4 from SK Hynix and Samsung
• NVL72 configuration: 72 Rubin GPUs and 36 Vera CPUs connected through NVLink 6 fabric, delivering 3.6 exaFLOPS of NVFP4 inference and 2.5 exaFLOPS of training
• 5x inference throughput improvement using NVIDIA's new 4-bit floating point (NVFP4) format

Huang framed the keynote around "AI as a Five-Layer Cake" — energy, chips, infrastructure, models, and applications. The first layer received unusual emphasis. Data centers already consume 2–3% of global electricity, and projections suggest that share could triple by 2030 as AI workloads scale. Huang highlighted renewable energy partnerships, including digital twins for ocean wave power generation, signaling that compute supply is no longer just a silicon problem — it's an energy problem.

Initial Vera Rubin samples are expected to ship to tier-one cloud providers by late 2026, with full production in early 2027. The next architecture, codenamed Feynman, is already on the roadmap for 2027.

The Supply Crisis No One Can Engineer Around​

While Vera Rubin's specifications grabbed headlines, the underlying supply story tells a more urgent tale. CEOs from TSMC, SK Hynix, Micron, Intel, NVIDIA, and Samsung have all delivered the same message: demand for advanced nodes, advanced packaging, and HBM is rising far faster than capacity can be built.

The bottleneck is comprehensive:

• HBM memory: SK Hynix confirmed "our entire 2026 HBM supply is sold out." Micron can meet only 55–60% of core customer demand. Samsung and SK Hynix have raised HBM3E prices by nearly 20% for 2026 contracts.
• Advanced packaging: TSMC's CoWoS (Chip-on-Wafer-on-Substrate) capacity — critical for assembling HBM stacks onto GPU packages — remains sold out through 2026.
• GPU allocation: Hyperscalers like Google, Microsoft, Amazon, and Meta have locked in multi-year allocations. Smaller enterprises face 36–52 week lead times, effectively locking them out of frontier AI hardware until 2027 or later.

The result is a two-tier compute market. A handful of hyperscalers command the vast majority of next-generation GPU capacity, while everyone else — startups, mid-market enterprises, research institutions, and sovereign AI initiatives — scrambles for whatever remains.

DePIN's Moment: From Fringe to Frontier​

This is where decentralized physical infrastructure networks enter the picture. While no DePIN network can manufacture NVIDIA GPUs out of thin air, these networks solve a different but equally critical problem: mobilizing the enormous pool of underutilized GPU capacity that already exists worldwide.

The DePIN compute sector has grown from $5.2 billion to over $19 billion in market capitalization within a single year, and the growth is backed by real usage metrics, not just token speculation.

Render Network has surpassed $2 billion in market cap after expanding from GPU rendering into AI inference workloads. Its launch of Dispersed — a dedicated subnet for AI workloads — positions the network at the intersection of creative and AI compute. Render delivers GPU rendering at up to 85% savings compared to AWS or Google Cloud.

Aethir reported nearly $40 million in quarterly revenue and over 1.4 billion compute hours delivered in 2025, serving 150+ enterprise clients. This isn't a testnet demo. It's production infrastructure generating real revenue.

io.net and Nosana each achieved market capitalizations exceeding $400 million during their growth cycles, aggregating idle GPU capacity from data centers, crypto miners, and consumer hardware into on-demand compute pools.

The pricing differential is striking. An NVIDIA H100 on a DePIN marketplace can cost 18–30x less than on AWS for comparable workloads. Even accounting for the reliability variance that forces some overprovisioning, DePIN networks offer 50–75% cost savings for batch workloads, inference tasks, and short-duration training runs.

The Enterprise Calculus Shifts​

Enterprise adoption of DePIN compute is following a predictable but accelerating pattern. The biggest blockers have been orchestration complexity, debugging distributed failures, lack of enforceable SLAs, and crypto-native procurement workflows that enterprise IT departments struggle to integrate.

But 2026 is changing the calculus. With centralized GPU access effectively rationed, enterprises are increasingly adopting hybrid architectures:

• Sensitive, low-latency models run locally on edge devices
• Massive training jobs stay with hyperscalers who have secured GPU allocations
• Flexible, burst-capacity inference routes to decentralized networks for cost arbitrage

This hybrid model turns DePIN from "interesting experiment" to "pragmatic overflow valve." When your AWS GPU quota is exhausted and NVIDIA's waitlist stretches past your product deadline, a 50% cost savings on a decentralized network stops being a philosophical choice about decentralization and becomes a business necessity.

The World Economic Forum's projection of a $3.5 trillion DePIN market by 2028 implies an extraordinary growth rate. Even at half that pace, DePIN would represent one of the fastest-growing infrastructure sectors in any industry.

Energy: The Hidden Bottleneck Behind the Chip Bottleneck​

Huang's emphasis on energy at GTC 2026 wasn't accidental. AI's electricity appetite is growing faster than the semiconductor supply chain can address. Current data center electricity consumption sits at 2–3% of global output, but projections suggest AI workloads alone could push this to 6–9% by 2030.

This energy bottleneck creates another structural advantage for DePIN networks. Centralized hyperscalers must build massive data centers in locations with abundant, affordable power — a process that takes 2–4 years from planning to operation. DePIN networks, by contrast, aggregate existing hardware in existing locations with existing power connections. The infrastructure is already plugged in.

Projects at the intersection of DePIN and energy, such as decentralized virtual power plants and tokenized renewable energy credits, are positioning to serve both sides of the equation: providing compute capacity while also coordinating the distributed energy resources needed to power it.

What Comes Next​

The Vera Rubin era will define AI infrastructure for the next two to three years. But the hardware that matters most isn't just what NVIDIA ships in 2027 — it's the millions of GPUs already deployed worldwide that sit idle for significant portions of each day.

Three dynamics will shape the next 12 months:

1. GPU scarcity intensifies before it eases. Vera Rubin production won't reach volume until early 2027. The current Blackwell generation remains supply-constrained. DePIN networks capturing overflow demand during this gap have a window to prove enterprise reliability at scale.

2. Hybrid compute architectures become standard. The binary choice between "hyperscaler or nothing" is dissolving. Enterprises will increasingly split workloads across centralized, edge, and decentralized infrastructure based on latency, cost, and availability requirements.

3. Energy becomes the binding constraint. Even when chip supply eventually loosens, power availability may not. DePIN's distributed model — inherently spread across diverse energy sources and geographies — provides structural resilience against localized power constraints that centralized data centers cannot match.

The irony of NVIDIA's GTC 2026 may be that its most important revelation wasn't Vera Rubin's breathtaking specifications. It was the confirmation that centralized AI infrastructure, no matter how powerful, faces physical limits that no amount of engineering can immediately solve. For the decentralized compute networks quietly aggregating the world's idle GPUs, those limits are an open door.

---

BlockEden.xyz provides high-performance RPC and API infrastructure for blockchain networks powering the next generation of decentralized applications — including the DePIN protocols building tomorrow's compute layer. Explore our API marketplace to start building.

A 72-billion-parameter language model trained entirely on commodity hardware, with no centralized cluster, no whitelist, and no corporate gatekeeper. That is what Bittensor's Subnet 3 delivered on March 10, 2026 — and the market noticed. TAO surged 56% in a single week while Render topped 40% gains as institutional capital rotated decisively into decentralized AI infrastructure.

The message from the market is unmistakable: DePIN is no longer a whitepaper narrative. It is generating real revenue, attracting institutional products, and challenging the cloud computing oligopoly at its most profitable frontier — artificial intelligence.

When the U.S. Department of Justice dismantled a $160 million smuggling ring moving NVIDIA chips to China in early 2026, it exposed a fundamental truth: centralized GPU supply chains are chokepoints — and chokepoints attract both enforcement and evasion. Meanwhile, a decentralized GPU cloud spanning 94 countries and 440,000+ containers was quietly rendering the entire debate less relevant.

Aethir, the largest decentralized physical infrastructure network (DePIN) for compute, has built something that neither AWS nor smuggling rings can replicate: a globally distributed GPU fabric where the nearest available H100 is routed to the client that needs it, regardless of which government controls the data center it sits in.

When Jensen Huang declared at CES 2026 that "the ChatGPT moment for physical AI is here," he was describing machines that understand, reason, and act in the real world. What he didn't say — but what a growing ecosystem of blockchain projects is betting on — is that those machines will also need to earn, spend, and own assets autonomously. Welcome to the era of DePAI: Decentralized Physical AI.

Aethir quietly crossed $127 million in annual revenue in 2025. Not in token emissions. Not in speculative incentive programs. In actual enterprise spending on GPU compute. That single data point may mark the moment decentralized compute stopped being a crypto experiment and started becoming a cloud business.

For years, the knock against Decentralized Physical Infrastructure Networks (DePIN) was simple: their economics ran on token printing, not customer invoices. Providers earned rewards denominated in volatile native tokens, demand was often synthetic, and the gap between "network activity" and "revenue" could be measured in orders of magnitude. But across 2025 and into early 2026, the leading GPU compute networks — Akash, io.net, Aethir, and Render — have been executing a pivot that the broader market hasn't fully priced in: the shift from token-subsidized supply to demand-driven cash flow.

Sixty-eight percent of all staked SOL sits in European data centers. That single statistic captures a vulnerability most Solana users never think about -- until a regional outage, a regulatory crackdown, or a fiber cut turns a theoretical risk into a live-fire crisis. On March 9, 2026, DoubleZero launched Phase II of its Delegation Program, redirecting 2.4 million SOL -- roughly $320 million at current prices -- toward validators in Sao Paulo, Singapore, Hong Kong, and Tokyo. The move is the most aggressive geographic rebalancing effort in any major proof-of-stake network's history, and it raises a question the entire industry should be asking: can economic incentives fix a decentralization problem that market forces created in the first place?

Every time an AI team retrieves a training dataset from AWS S3, a quiet tax gets added to the bill. It is called an egress fee, and across the cloud industry it silently inflates storage costs by 30-80%, turning what looks like affordable object storage into a budgetary black hole. In March 2026, a startup called Akave launched its answer: an S3-compatible decentralized storage platform with flat-rate pricing, zero egress fees, and cryptographic proof that your data actually exists where it is supposed to.

Backed by $6.65 million from Protocol Labs, the Avalanche Foundation, the Filecoin Foundation, Big Brain Holdings, and others, Akave Cloud is not just another Web3 storage experiment. It is a production-grade infrastructure play targeting the fastest-growing segment of cloud spending: AI data lakes.

The three highest-revenue DePIN projects in 2026 share one thing in common: they all sell GPU compute to AI companies. Not storage. Not wireless bandwidth. Not sensor data. Compute — the single most constrained resource in the global technology stack.

That fact alone tells you everything about where Decentralized Physical Infrastructure Networks have landed after years of searching for product-market fit. The sector that once ran on token incentives and speculative flywheel economics now generates real revenue from the most demanding buyers in tech: AI model developers who need GPUs yesterday.

When robots start earning their own paychecks, who controls their wallets? That's the trillion-dollar question driving DePAI—Decentralized Physical AI—a paradigm shift that's moving physical robots and AI systems from corporate data centers to community-owned infrastructure. While Web3 has spent years promising to decentralize the digital world, 2026 marks the year this vision collides with the physical realm: autonomous vehicles, humanoid robots, and AI-powered IoT devices operating on blockchain rails.

The numbers tell a compelling story. The World Economic Forum projects the DePIN (Decentralized Physical Infrastructure Networks) market will explode from $20 billion today to $3.5 trillion by 2028—a staggering 6,000% increase. What's driving this growth? The convergence of AI and blockchain is creating what industry insiders now call "DePAI"—infrastructure that enables distributed machine learning, autonomous economic agents, and community-owned robotics networks at unprecedented scale.

This isn't speculative tokenomics anymore. Real revenue is flowing through decentralized networks: Aethir posted $166 million in annualized revenue serving 150+ enterprise AI clients, Helium's decentralized wireless network hit $13.3 million in annualized revenue through partnerships with T-Mobile and AT&T, and Grass is generating approximately $33-85 million annually selling web-scraped data to AI companies. The shift from "token speculation" to "business revenue models" has arrived.

From DePIN to DePAI: The Evolution of Decentralized Infrastructure​

To understand DePAI, you need to grasp its foundation: DePIN (Decentralized Physical Infrastructure Networks). DePIN uses blockchain and token incentives to crowdsource physical infrastructure—wireless networks, GPU compute, storage, sensors—that traditionally required massive capital expenditure from corporations. Think Uber, but for infrastructure: individuals contribute resources (bandwidth, GPUs, storage) and earn tokens in return.

DePAI takes this concept further by adding autonomous AI agents into the mix. It's not just about decentralizing infrastructure ownership—it's about enabling AI systems and physical robots to interact with that infrastructure autonomously, transact in decentralized markets, and execute complex tasks without centralized cloud dependencies.

The seven-layer DePAI stack illustrates this evolution:

1. AI Agents - Autonomous software entities that make decisions and execute transactions
2. Robotics - Physical embodiments (humanoid robots, drones, autonomous vehicles)
3. Decentralized Data Streams - Real-time sensor data, location data, environmental inputs
4. Spatial Intelligence - Mapping, navigation, and environmental understanding
5. Infrastructure Networks - DePIN for compute, storage, connectivity
6. The Machine Economy - Peer-to-peer markets where machines transact directly
7. DePAI DAOs - Governance layers enabling community ownership and decision-making

This stack transforms robots from isolated corporate assets into economically autonomous actors in a decentralized ecosystem. Imagine a delivery drone that autonomously books GPU compute for route optimization, purchases bandwidth access through a DePIN marketplace, and settles payments via smart contracts—all without human intervention.

The Enterprise Revenue Breakout: Aethir's $166M Lesson​

For years, DePIN projects struggled with the "chicken-and-egg" problem: how do you bootstrap supply (people contributing resources) without demand (paying customers), and vice versa? Aethir cracked this problem with a laser focus on enterprise clients rather than retail speculators.

In Q3 2025 alone, Aethir generated $39.8 million in revenue, reaching a $147+ million annual recurring revenue (ARR) run rate. By early 2026, this figure hit $166 million ARR. The key differentiator? These revenues came from 150+ enterprise clients across AI, gaming, and Web3—not from token emissions or subsidies.

With over 435,000 enterprise-grade GPUs distributed across 200+ locations in 93 countries, Aethir provides more than $400 million worth of compute capacity while maintaining an exceptional 98.92% uptime. That's infrastructure reliability comparable to AWS or Google Cloud, but delivered through a decentralized network where GPU owners earn yield and customers pay 50-85% less than hyperscaler prices.

The business model is straightforward: AI companies need massive compute for training and inference. Centralized cloud providers like AWS charge premium rates and face GPU scarcity (SK Hynix and Micron have announced their entire 2026 output is sold out). Aethir aggregates idle GPU capacity from data centers, mining operations, and enterprise partners, making it available through a decentralized marketplace at fractional costs.

For 2026, Aethir is doubling down on agentic AI—enabling autonomous AI agents to book, pay for, and optimize GPU usage in real-time without human operators. This positions DePAI infrastructure not just as a cost-efficient alternative to centralized cloud, but as the native rails for the emerging machine economy.

Helium's Hybrid Model: Carrier Offload Meets Community Networks​

While Aethir focuses on compute, Helium tackles connectivity. What started in 2019 as a community-driven IoT network has evolved into a full-stack wireless DePIN supporting both IoT and 5G mobile services. By Q3 2025, the Helium Network had transferred over 5,452 terabytes of data offloaded from major U.S. mobile carriers, representing significant quarter-over-quarter growth.

The "carrier offload" model is where DePAI meets real-world telecommunications. Major carriers like T-Mobile, AT&T, Movistar, and Google Orion partner with Helium to offload customer data to community-run hotspots in high-traffic urban areas. The carrier pays the network a fee, and that revenue flows to hotspot operators who provide the physical infrastructure.

Despite some confusion in media reports, Helium does not have a formal carrier offload agreement directly with T-Mobile as a telecom-to-telecom partnership. Instead, T-Mobile subscribers can connect to Helium's network at select locations through third-party arrangements, and carriers benefit from reduced congestion by offloading traffic to Helium's 26,000+ Wi-Fi sites.

Helium Mobile, the network's MVNO (Mobile Virtual Network Operator) service, exemplifies the "Hybrid MNO" model: users get unlimited mobile plans for $20/month by seamlessly switching between Helium's community network and T-Mobile's backbone. When you're near a Helium hotspot, your traffic gets routed through DePIN infrastructure. When you're not, T-Mobile's network serves as backup.

This hybrid approach proves DePAI doesn't need to replace centralized infrastructure entirely—it can augment it, capturing high-margin use cases (urban density, IoT sensors, stationary devices) while leaving low-margin scenarios to traditional providers. The result: $13.3 million in annualized revenue for a network bootstrapped by retail participants, not telecom giants.

Grass: Monetizing Idle Bandwidth for AI Training Data​

If Aethir is selling compute and Helium is selling connectivity, Grass is selling data—specifically, web data scraped by a decentralized network of 2.5 million+ users who contribute their unused internet bandwidth.

AI companies face a critical bottleneck: they need massive, diverse datasets to train large language models (LLMs), but scraping the public web at scale requires enormous bandwidth and IP diversity to avoid rate limits and geographic blocks. Grass solved this by crowdsourcing bandwidth from everyday internet users, turning their home connections into a distributed web-scraping network.

The revenue model is straightforward: AI labs purchase structured datasets through the Grass network for model training, paying the Grass Foundation in fiat or crypto. The GRASS token serves as the "primary vehicle for value accrual," distributing revenue back to node operators and stakers who provide the underlying infrastructure.

While exact revenue figures vary across sources, Grass monetizes less than 1% of its 2.5M+ user base and already generates substantial early revenue estimates ranging from $33 million to $85 million annually. The founder casually mentioned a "mid-8 figure revenue" in a recent demo, suggesting the network is generating $50+ million per year. With 8.5 million monthly active users and growing commercial deals with AI labs, Grass is scaling network capacity for both training datasets and live context retrieval data to serve AI clients through 2026-2027.

What makes Grass a DePAI case study rather than just a data marketplace? The network enables autonomous AI agents to access real-time, decentralized web data without relying on centralized APIs that can be censored, rate-limited, or shut down. As AI agents become more autonomous and economically active, they'll need infrastructure that's as permissionless and decentralized as they are.

The Robotics Revolution: When Machines Need DePAI Infrastructure​

DePAI's ultimate vision extends beyond compute, connectivity, and data—it's about enabling physical robots to operate as autonomous economic agents. Morgan Stanley analysts predict the humanoid robotics industry could generate up to $4.7 trillion in annual revenue by 2050. But here's the critical question: will these robots be controlled by a handful of corporations (Boston Dynamics under Hyundai, Tesla's Optimus, Google's robotics division), or will they operate on decentralized infrastructure owned by communities?

Projects like peaq, XMAQUINA, and elizaOS are pioneering the DePAI approach to robotics:

• peaq functions as the "Machine Economy operating system," enabling robots, sensors, and IoT devices to interact via self-sovereign IDs, transact peer-to-peer, and offer data and services through decentralized marketplaces. Think of it as the Ethereum for machines.

• XMAQUINA advances DePAI through a DAO structure, giving a global community liquid exposure to leading private robotics companies developing next-generation humanoids. Instead of robots being corporate assets, investors pool resources and democratize ownership in robotics companies via blockchain-based governance.

• elizaOS bridges decentralized AI agents and robotics by turning autonomous intelligence into real-world workflows. It extends naturally into robotics where systems must process data locally and coordinate tasks without relying on fragile centralized clouds.

The core idea is "universal basic ownership" as an alternative to universal basic income (UBI). If robots displace human labor at scale, DePAI offers a model where everyday people profit from machine labor as owners and stakeholders in the networks, not just passive recipients of government transfers.

By 2030, industry forecasts suggest more than half of all AI-driven robots will run workloads on decentralized GPU networks like Aethir, not on AWS, Azure, or Google Cloud. They'll use DePIN wireless networks like Helium for connectivity, access real-time data through networks like Grass, and settle transactions via smart contracts. The vision is a machine economy where autonomous agents and physical robots interact in permissionless markets, owned and governed by DAOs rather than monopolies.

Why 2026 Marks the Shift from Speculation to Revenue​

For years, DePIN and Web3 infrastructure projects were funded by token emissions and venture capital, not paying customers. That model worked during bull markets but collapsed spectacularly when crypto entered bear markets. Projects with no real revenue but high token inflation saw their networks and valuations evaporate.

2026 marks a paradigm shift. The metrics that matter now are:

• Network revenue - How much fiat or stablecoin revenue is the network generating from actual customers?
• Utilization rates - What percentage of the network's capacity is being actively used by paying users?
• Enterprise adoption - Are real businesses (not just crypto-native protocols) using the infrastructure?

Aethir, Helium, and Grass demonstrate this shift in action:

• Aethir's $166M ARR comes from 150+ enterprise clients, not token incentives.
• Helium's $13.3M annual revenue comes from carrier offload partnerships and MVNO subscribers, not speculative hotspot purchases.
• Grass's $33-85M revenue comes from AI companies buying datasets, not airdrop farmers.

The GPU-as-a-service market alone is estimated to be worth $35-70 billion by 2030, with accelerated compute workloads growing at more than 30% CAGR. Decentralized services are competing on cost (50-85% savings vs. AWS/GCP), flexibility (global distribution, no vendor lock-in), and resistance to centralized control—values that resonate especially with AI developers concerned about censorship and platform risk.

Compare this to traditional DePIN tokens that collapsed when incentives dried up. The difference is sustainable unit economics: if the network earns more revenue from customers than it spends on token emissions and operations, it can survive indefinitely without bull market bailouts.

The $3.5 Trillion Question: Can DePAI Actually Scale?​

The World Economic Forum's $3.5 trillion projection by 2028 sounds audacious, but it hinges on three critical factors:

1. Regulatory Clarity

Physical infrastructure—wireless networks, data centers, transportation systems—operates under heavy regulation. Can DePIN and DePAI networks navigate telecom licensing, data privacy laws (GDPR, CCPA), and robotics safety standards while maintaining decentralization? Helium's carrier partnerships suggest yes, but regulatory risk remains high.

2. Enterprise Adoption

AI companies and robotics firms need infrastructure that's reliable, compliant, and cost-effective. Aethir's 98.92% uptime and enterprise-grade SLAs prove decentralized networks can compete on reliability. But will Fortune 500 companies trust critical workloads to community-owned infrastructure? The next 12-24 months will be telling.

3. Technological Maturation

DePAI requires seamless integration across blockchain (payments, identity, governance), AI (autonomous agents, machine learning), and physical systems (robotics, sensors, edge compute). Many pieces still need interoperability standards, better developer tools, and reduced latency for real-time applications.

The bullish case is compelling: global AI infrastructure spending is projected to hit $5-8 trillion through 2030, and decentralized networks are capturing an increasing share by offering cost, flexibility, and sovereignty advantages. The bearish case warns of centralization creep (a few large node operators dominating networks), regulatory crackdowns, and competition from hyperscalers who could match DePIN pricing through economies of scale.

What Comes Next: The Machine Economy Goes Live​

As we move deeper into 2026, several trends will accelerate DePAI's evolution:

Agentic AI proliferation - AI agents are moving from chatbots to autonomous economic actors. They'll need DePAI infrastructure for permissionless access to compute, data, and connectivity.

Open-source model adoption - As more companies run open-source LLMs (Llama, Mistral, etc.) instead of relying on OpenAI/Anthropic APIs, demand for decentralized inference will surge.

Robotics commercialization - Humanoid robots entering warehouses, factories, and service industries will need decentralized infrastructure to avoid vendor lock-in and enable interoperability.

Tokenized incentives for edge nodes - The next wave of DePIN projects will focus on edge compute (processing data close to where it's generated) rather than centralized data centers. This fits perfectly with latency-sensitive robotics and IoT applications.

For developers and investors, the playbook is shifting: look for projects with real revenue, sustainable unit economics, and enterprise traction. Avoid networks sustained purely by token emissions or speculative NFT sales. The DePAI winners will be those bridging Web3's permissionless ethos with the reliability and compliance standards enterprise customers demand.

For builders developing AI applications that require reliable, cost-efficient infrastructure, BlockEden.xyz offers enterprise-grade API access to leading blockchain networks. Explore our services to build on infrastructure designed for the decentralized future.

Sources​

• DePAI Revolution: Powering Physical AI and Robotics
• Understanding DePAI | Messari
• Aethir's 12-Month Strategic Roadmap
• How Aethir Became the Top Enterprise DePIN Compute Platform
• State of Helium Q3 2025 | Messari
• Helium's carrier-grade turn: AT&T, Movistar, and free mobile
• Research Report: Grass — The Expanding AI Data Bank
• Reshaping the data layer of AI with Grass - Blockworks
• What robots need to operate as Decentralized Physical AI (DePAI) - peaq
• The Rise of DePIN - BlockEden.xyz
• Decentralized GPU Networks 2026 - BlockEden.xyz