Fluence Plans $80 Billion AI Compute Layer to Challenge Cloud Giants

Fluence is constructing a compute layer that is open, low-cost, and enterprise-grade, which centralized clouds cannot achieve. This layer is designed to be sovereign, transparent, and accessible to everyone. The year 2025 has begun with cloud giants aggressively investing in AI infrastructure, aiming to dominate the market. Microsoft is investing over $80 billion in new data centers, Google has launched its AI Hypercomputer, Oracle is investing $25 billion into its Stargate AI clusters, and AWS is prioritizing AI-native services. Specialized players are also scaling rapidly, with CoreWeave raising $1.5 billion in its March IPO and currently valued at over $70 billion.

As AI becomes critical infrastructure, access to compute power will be a defining battle. While hyperscalers consolidate and centralize compute power by building exclusive data centers and vertically integrating silicon, Fluence offers a different vision—a decentralized, open, and neutral platform for AI compute. This platform tokenizes compute to meet AI’s exponential demand, with FLT serving as a Real-World Asset (RWA) tokenized compute asset. Fluence is already collaborating with top decentralized infrastructure networks across AI and storage on multiple initiatives, reinforcing its position as a neutral compute-data layer.

Fluence's roadmap for 2025–2026 focuses on three key action areas: launching a global GPU-powered compute layer, hosting AI models and unified inference, and enabling verifiable, community-driven service-level agreements (SLAs). The first initiative involves supporting GPU nodes globally, enabling compute providers to contribute AI-ready hardware. This will upgrade the Fluence platform from CPU-based capacity to an additional AI-grade compute layer designed for inference, fine-tuning, and model serving. Fluence will integrate container support for secure, portable GPU job execution, enabling reliable ML workload serving and establishing critical infrastructure for future applications across the decentralized network.

Fluence will also explore privacy-preserving inference through confidential computing for GPUs, keeping sensitive data private while reducing AI inference costs. Using trusted execution environments (TEE) and encrypted memory, this R&D initiative enables sensitive workload processing while maintaining decentralization and supporting sovereign agent development. Key milestones include GPU node onboarding in Q3 2025, GPU container runtime support live in Q4 2025, confidential GPU computing R&D track kickoff in Q4 2025, and pilot confidential job execution in Q2 2026.

The second initiative involves providing one-click deployment templates for popular open-source models, including LLMs, orchestration frameworks like LangChain, agentic stacks, and MCP servers. The Fluence platform AI stack will be expanded with an integrated inference layer for hosted models and agents, simplifying AI model deployment while leveraging community contributions and external development support. Key milestones include model and orchestration templates live in Q4 2025 and inference endpoints and routing infrastructure live in Q2 2026.

The third initiative introduces a new approach to network trust and resilience through Guardians—retail and institutional actors who verify compute availability. Guardians monitor infrastructure through decentralized telemetry and earn FLT rewards for enforcing SLAs. This turns an enterprise-grade infrastructure network into something anyone can participate in without needing to own hardware. The Guardian program is complemented by the Pointless Program, a gamified reputation system that rewards community contributions and leads to Guardian eligibility. Key milestones include the first batch of Guardians in Q3 2025 and the full rollout of Guardians and programmatic SLAs in Q4 2025.

Fluence is also integrating AI compute with a composable data stack, building deep integrations with decentralized storage networks to provide developers with access to verifiable datasets alongside execution environments. These integrations will allow users to define jobs that access persistent, distributed data and run on GPU-backed nodes, turning Fluence into a full-stack AI backend orchestrated via FLT. The network will offer composable templates and prebuilt SDK modules for connecting compute jobs with storage buckets or on-chain datasets. Developers building AI agents, LLM inference tools, or science applications will be able to treat Fluence like a modular AI pipeline with open data, compute, and validation stitched together by protocol logic. Key milestones include decentralized storage backups in Q1 2026 and integrated dataset access for AI workloads in Q3 2026.

With a roadmap focused on GPU onboarding, verifiable execution, and seamless data access, Fluence is laying the foundation for the next era of AI. This era will not be controlled by a handful of hyperscalers but powered by a global community of cooperating and decentralized compute providers and participants. The infrastructure for AI must reflect the values we want AI to serve: openness, collaboration, verifiability, and accountability. Fluence is turning that principle into a protocol.