On October 1, two of the largest moves in agent infrastructure shipped on the same day — and neither was primarily a technical announcement. DigitalOcean launched Agent Droplets, bundling microVMs, memory, session storage, inference on hosted models, and governed access to 16,000+ tools into three subscription tiers. Microsoft launched Agent 365 GCC, packaging agent discovery, identity, governance, and security into a single control plane for government cloud environments. The agent stack is no longer something you assemble. It is something you buy.
This follows the commoditization we documented in late September, when three vendors shipped identical harness architectures within 48 hours. AWS Strands, DigitalOcean Managed Agents, and Aiven Runtime all converged on the same pattern: shell access, file manipulation, web tools, long-term memory, and subtask delegation. OpenAI DevDay then validated the harness as the industry default. The technical implementation of agent orchestration stopped being a differentiator weeks ago. What changed on October 1 is how it gets sold.
The pricing models are bifurcating. AWS AgentCore still bills memory, gateways, and observability as separate metered line items — runtime alone costs $0.0895 per vCPU-hour. Specialized sandbox providers like Northflank, E2B, and Modal compete on per-second granularity and bring-your-own-cloud options. DigitalOcean’s Agent Droplets take the opposite approach: a $5 Free Trial, a $50 per month Pro plan with a 15% usage discount, and a $200 per month Team plan with a 20% discount. Unlimited agents, unlimited seats, enterprise IAM included. As DigitalOcean CPTO Vinay Kumar framed it: “Fourteen years ago a Droplet made the cloud something one developer could understand and afford. Agents are in the same place now.”
The bundling is a direct response to a deployment problem, not just a marketing choice. A typical agent stack today requires six vendors and dozens of pricing units — harness, sandbox, storage, inference, search, each billing in its own currency on a separate invoice. No one on the team can calculate what a single agent run cost. According to Cisco research, only 5% of agentic AI projects have reached broad production. The bottleneck is not the architecture. It is the operational overhead of running it.
Microsoft’s Agent 365 GCC solves a different bottleneck with the same bundled approach. By packaging agent discovery, Entra Agent ID identity, Purview data protection, Defender threat security, and Intune device management into a FedRAMP High-authorized control plane, Microsoft is selling compliance as infrastructure. The certifications — FedRAMP High, DFARS, ITAR, CJIS — took years to obtain and cannot be replicated by providers that lack existing government authorization. For federal agencies that cannot deploy commercial AI agents without FedRAMP approval, Agent 365 GCC eliminates 12 to 18 months of custom compliance work.
The deeper implication is that the competitive moat is moving. When the harness pattern was a thesis, differentiation came from architecture. Now that the harness is a commodity, differentiation comes from two places that resist easy replication: the intelligence layer and the compliance wrapper. DigitalOcean bundles inference on hosted open models, but frontier models like GPT and Claude remain excluded from the bundle discount, billed at list prices. Microsoft’s compliance certifications represent institutional investment that takes years to accumulate. Neither is something a new entrant can copy by shipping a better sandbox.
For builders evaluating infrastructure, the decision framework is shifting. Per-vCPU-hour optimization matters less than total cost of ownership and governance readiness. The question is no longer “which harness is fastest” but “which provider lets me get to production without rebuilding my compliance stack.” The harness pattern answered the first question. The bundled SKU era is answering the second.