Developer Tools & Platforms — 2026年7月1日 月次レポート
重要な発見
エグゼクティブサマリー（5件）
- •The developer tools market crossed a structural threshold in June 2026: agentic infrastructure is no longer a roadmap item but a production reality, validated by multiple independent customer deployments at extreme operational leverage ratios and anchored by Vercel's Agent Stack, Atlassian's Claude Agent for Jira, and Cloudflare's agent-native deployment primitives.
- •AI model economics bifurcated sharply across the month: DeepSeek V4's 17-point token share surge at near-zero spend share established a permanent low-cost tier for high-volume workloads, while Anthropic's growing spend dominance (65% by May) confirmed that frontier models are concentrating in quality-critical, high-consequence tasks — making multi-model routing a core architectural requirement.
- •Supabase's $500M Series F at a $10B valuation was the month's defining funding event, validating open-source backend infrastructure as a standalone market and compounding Supabase's distribution moat through sequential integration into both Claude and ChatGPT ecosystems.
- •Enterprise AI governance emerged as a critical gap: inference theft attacks, AI coding costs projected to exceed human developer costs, record vulnerability volumes on GitHub, and the finding that nearly two-thirds of organizations lack IT asset visibility to control AI costs collectively signal that security and cost governance infrastructure will be the next major product battleground.
- •The gap between individual AI productivity gains and enterprise business outcomes — 85% individual AI usage versus 29% team-level embedding per Atlassian's survey — combined with Adobe's deferred price increase signals that the frictionless AI monetization era is ending, forcing developer platform vendors to demonstrate measurable team-level ROI or face pricing pressure at renewal.
今回の要点（5件）
- 1.Agentic deployments on Vercel surpassed 30% of all weekly deployments — up 1,000% in six months — with production customers including Superset (~600 preview deployments/day), General Intelligence (4,000+ simultaneous branches), and Okara (4 billion daily tokens, 4-person team) providing concrete evidence that agentic infrastructure has moved from vision to measurable production scale.
- 2.Supabase raised a $500M Series F at a $10B pre-money valuation led by GIC on June 4, 2026, cementing open-source backend infrastructure as a standalone, investor-validated market category and providing Supabase with capital to deepen its position as the default database layer for agentic applications.
- 3.DeepSeek V4 captured 17% of Vercel AI Gateway token share in a single month while accounting for only ~1% of spend, confirming that low-cost capable models have cleared the production quality bar — while Anthropic simultaneously grew its spend share from 61% to 65%, holding 70–80% of spend across every high-stakes agentic use case.
- 4.Vercel launched the Agent Stack on June 17, 2026, consolidating previously separate agentic primitives into a single named product; Atlassian launched Claude Agent for Jira on June 24, 2026, making AI agents first-class assignable teammates; and Cloudflare introduced Temporary Accounts enabling agents to deploy Workers without human account setup — collectively signaling platform-wide convergence on agent-native developer tooling.
- 5.B2B SaaS pricing power showed its first major crack since 2022 (Adobe's deferred price increase per SaaStr), AI coding costs are projected to exceed human developer costs per Gartner analysis, and Atlassian's survey of 12,000+ workers found only 29% have embedded AI into team workflows despite 85% individual usage — establishing enterprise AI cost governance and adoption gap as the defining challenges for the next period.
市場動向
Agentic Infrastructure Transitions from Vision to Production-Validated Market Category
Across the month, the agentic infrastructure narrative accumulated compounding production evidence that moved it from strategic positioning to a measurable market reality. Vercel reported over 30% of weekly deployments initiated by coding agents — up 1,000% in six months — with Claude Code accounting for 75% of agent-driven deployments and agent-deployed projects 20x more likely to call AI inference providers. Three independent production customer case studies corroborated the platform statistic…
AI Model Economics Bifurcate: Low-Cost Models Capture Volume, Frontier Models Capture Value
The month's AI Gateway production data revealed a structural bifurcation in how production teams allocate AI spend. DeepSeek V4 captured 17% of Vercel AI Gateway token share in a single month following its launch — rising from under 1% — while its spend share remained near 1%, reflecting pricing of $0.14 input / $0.28 output per million tokens, roughly 20–50x lower than comparable Anthropic models. DeepSeek drove 49% of coding agent token volume in May while accounting for only 4% of cost. Simul…
Inference Theft and AI Security Costs Emerge as Structural Platform Risks
A new and economically distinct security threat category crystallized across the month. Vercel documented an inference theft attack on April 12, 2026 in which traffic to its docs AI chat endpoint spiked to 1,300 requests per minute — roughly 10x normal volume — representing an inference cost run rate of over $10,000 per day, delivered through residential proxies that defeated standard per-IP rate limits. The asymmetric economics are stark: a single frontier model prompt costs approximately $2 ve…
Enterprise AI Adoption Gap Widens Between Individual Productivity and Business Outcomes
Multiple data points across the month converged on a consistent finding: individual AI productivity gains are not translating into enterprise-level business results. Atlassian's State of Teams 2026 survey of 12,000+ knowledge workers found that while 85% of workers use AI, only 29% have embedded it into team workflows, and business results are not catching up to individual productivity gains. Atlassian's Teamwork Lab research found that workers who disclose AI use are judged 10x lazier than peer…
B2B SaaS Pricing Power Under Pressure as AI Cost Scrutiny Intensifies
The month produced converging signals that the frictionless AI monetization era for SaaS vendors is ending. SaaStr reported that Adobe deferred a significant annual price increase — characterized as the first major crack in B2B pricing power since 2022 — with the prior four-year playbook of raising prices every 12–18 months and adding AI SKUs as additional charges now under customer pressure. CIO Dive reported that AI coding costs are projected to soon exceed human developer costs per Gartner an…
Open-Source Backend Infrastructure Validated as Standalone Investment Category
Supabase's $500M Series F at a $10B pre-money valuation led by GIC, announced June 4, 2026, was the month's defining funding event and validated open-source backend infrastructure as a standalone, investor-recognized market category. The funding followed a sustained period of product and ecosystem expansion: Supabase became an official ChatGPT app on May 8, 2026, following its February 2026 Claude connector integration; achieved ISO 27001 certification; launched Multigres v0.1 Alpha described as…
AI Governance Migrates from Application Layer to Infrastructure Enforcement
Across the month, enterprise AI governance requirements — covering both cost control and security monitoring — consistently moved from application-level implementations toward infrastructure-layer enforcement. Cloudflare announced real-time spend limits in AI Gateway on June 5, 2026, integrated with Cloudflare Access for identity-driven cost governance at the user or team level. AWS added CloudWatch metrics for the bedrock-mantle endpoint on June 1, 2026, covering inference counts, input and out…
競合動向
Vercel Consolidates Agentic Platform Leadership with Agent Stack and Flags
Vercel's competitive trajectory across the month moved from individual agentic primitives to a consolidated named product. The Agent Stack, launched June 17, 2026, unified AI SDK, AI Gateway, Sandboxes, Fluid Compute, Workflows, and Observability into a single offering for production-grade agents. Vercel Flags, launched June 23, 2026, extended the platform beyond deployment into release management with server-side evaluation, automatic flag registration, and a CLI accessible to both developers a…
Supabase Compounds Distribution Moat Through Funding, AI Ecosystem Integration, and Compliance
Supabase's competitive position strengthened across multiple dimensions during the month. The $500M Series F at a $10B pre-money valuation provided capital to accelerate its open-source backend infrastructure strategy. Sequential integration into both Claude (February 2026) and ChatGPT (May 8, 2026) established AI assistant ecosystems as a primary distribution channel, creating discovery advantages as developers increasingly interact with infrastructure through conversational interfaces. ISO 270…
Atlassian Evolves from AI Feature Vendor to Agent Orchestration Platform
Atlassian's competitive positioning shifted meaningfully across the month from AI-enhanced project management toward agent orchestration infrastructure. The introduction of Cursor in Jira enabled work items to be assigned directly to Cursor as an AI coding agent. Bitbucket Agentic Pipelines added Claude Code support. The Claude Agent for Jira launch on June 24, 2026 made AI agents first-class assignable and mentionable teammates within existing Jira workflows. @Jira for Slack, trusted by 3.5 mil…
AWS Bedrock Executes API Compatibility Strategy to Capture OpenAI and Anthropic Developer Base
AWS pursued a deliberate strategy across the month to eliminate switching friction for developers already invested in OpenAI or Anthropic tooling. The redesigned Amazon Bedrock console, launched June 4, 2026, supports the OpenAI Responses API, OpenAI Chat Completions API, and Anthropic Messages API via the bedrock-mantle endpoint, enabling existing applications to run on Bedrock with minimal code changes. CloudWatch metrics for the bedrock-mantle endpoint, added June 1, 2026, cover inference cou…
Cloudflare Builds Agent-Native Infrastructure Primitives Across the Month
Cloudflare shipped a sequence of capabilities across the month that collectively position it as infrastructure for machine-driven rather than human-driven workflows. Real-time spend limits in AI Gateway, announced June 5, 2026 and integrated with Cloudflare Access for identity-driven cost governance, addressed the enterprise AI cost control problem at the network layer. Temporary Accounts on Cloudflare Workers, announced June 19, 2026, enabled any AI agent to deploy a live Worker in seconds with…
GitHub Faces Dual Pressure: Record Security Workload and Agentic Era Positioning
GitHub's competitive trajectory across the month was defined by two converging pressures. On the strategic positioning side, GitHub Universe 2026 was announced for October 28–29, 2026 at Fort Mason Center in San Francisco with an explicit 'agentic era' theme, reinforcing GitHub's commitment to positioning its platform as core infrastructure for agentic software development. On the operational side, GitHub reported that the Advisory Database is processing more vulnerability reports than ever befo…
制度・規制動向
Inference Theft Establishes Per-Request Verification as a New Security Compliance Requirement
Vercel's documentation of the April 12, 2026 inference theft attack — in which residential proxies drove traffic to 1,300 requests per minute at a projected cost run rate of over $10,000 per day — established a new security compliance category for developer platform operators. The attack exploited the fundamental asymmetry between AI inference costs (approximately $2 per frontier model call) and standard HTTP request costs (approximately $2 per million), creating high-margin incentives for attac…
Enterprise AI Cost Governance Moves to Infrastructure-Layer Enforcement
Across the month, enterprise AI cost governance requirements consistently migrated from application-level controls to infrastructure-layer enforcement. Cloudflare's real-time spend limits in AI Gateway, integrated with Cloudflare Access for identity-driven budgets and policies at the user or team level, represent the clearest example of this pattern. AWS's CloudWatch metrics for the bedrock-mantle endpoint provided granular cost attribution at account, project, model, and project-and-model level…
AWS GovCloud AI Authorizations Open Regulated Public Sector Markets to Developer Tooling
AWS announced on June 25, 2026 that OpenAI GPT, OpenAI GPT OSS, and NVIDIA Nemotron models on Amazon Bedrock received FedRAMP High and DoD IL-4/5 authorization in AWS GovCloud (US). Kiro also achieved FedRAMP High and DoD IL-4/5 authorization in AWS GovCloud (US) on the same date. These authorizations represent the first time AI-powered developer tools have been qualified for the most sensitive government and defense environments at IL-4/5 sensitivity levels, expanding the addressable market for…
Supabase Formalizes Enterprise Security Posture with ISO 27001 and Supply Chain Guidance
Supabase made two notable compliance and security moves during the period. The company announced ISO 27001 certification as of April 22, 2026, a formal information security management standard relevant to enterprise developer platform procurement. On May 26, 2026, Supabase published guidance on protecting projects from npm supply chain attacks, outlining practical steps for developers to reduce risk from compromised open-source packages. These developments reflect a broader trend of developer in…
ソース活動
先月からの変化
Supabase $500M Series F at $10B Valuation
Supabase announced a $500M Series F at a $10B pre-money valuation led by GIC on June 4, 2026, validating open-source backend infrastructure as a standalone investment category and providing capital to accelerate its AI ecosystem integration strategy. The company had previously become an official ChatGPT app (May 8, 2026), achieved ISO 27001 certification, launched Multigres v0.1 Alpha, and reached 100,000 GitHub stars.
Vercel Agent Stack Launch
Vercel launched the Agent Stack on June 17, 2026, consolidating previously separate agentic primitives — AI SDK, AI Gateway, Sandboxes, Fluid Compute, Workflows, and Observability — into a single named product offering for production-grade agents. This marked the transition from agentic infrastructure as a set of individual primitives to a coherent named product category.
Vercel Flags: Platform-Native Feature Flag System
Vercel launched Vercel Flags on June 23, 2026, a platform-native feature flag system with server-side evaluation, automatic flag registration, and a CLI accessible to both developers and coding agents. The v0 team runs hundreds of flags at any given moment including for AI model routing. This extends Vercel's platform beyond deployment into release management, closing the last major human-gated step in the agent-driven deployment loop.
Atlassian Claude Agent for Jira Launch
Atlassian launched Claude Agent for Jira on June 24, 2026, making AI agents first-class assignable and mentionable teammates within existing Jira workflows. Combined with Cursor integration in Jira, Bitbucket Agentic Pipelines support for Claude Code, and @Jira for Slack, this positions Jira as an agent orchestration layer for enterprise software development teams.
AWS Bedrock OpenAI and Anthropic API Compatibility Layer
AWS launched a redesigned Amazon Bedrock console on June 4, 2026 supporting the OpenAI Responses API, OpenAI Chat Completions API, and Anthropic Messages API via the bedrock-mantle endpoint, enabling existing applications to migrate to Bedrock with minimal code changes. CloudWatch metrics for the bedrock-mantle endpoint were added June 1, 2026. AWS GovCloud received FedRAMP High and DoD IL-4/5 authorization for OpenAI GPT, NVIDIA Nemotron, and Kiro on June 25, 2026.
DeepSeek V4 Captures 17% of AI Gateway Token Share
DeepSeek V4 captured 17% of Vercel AI Gateway token share in a single month following its launch, rising from under 1%, while spend share remained near 1%. DeepSeek drove 49% of coding agent token volume in May 2026 while accounting for only 4% of cost. Priced at $0.14 input / $0.28 output per million tokens — roughly 20–50x lower than comparable Anthropic models — this confirmed that low-cost capable models have cleared the production quality bar for high-volume workloads.
Cloudflare Agent-Native Infrastructure: Temporary Accounts and Saga Rollbacks
Cloudflare announced Temporary Accounts on Cloudflare Workers on June 19, 2026, enabling any AI agent to deploy a live Worker in seconds without a pre-existing human account. On June 29, 2026, Cloudflare shipped saga-style rollbacks for Cloudflare Workflows, allowing developers to specify compensating actions for each step of long-running multi-step applications. Together these represent a new category of infrastructure designed natively for machine-driven workflows.
Inference Theft Attack Vector Documented and Mitigated
Vercel documented an inference theft attack on April 12, 2026 in which traffic spiked to 1,300 requests per minute at roughly 10x normal volume, representing an inference cost run rate of over $10,000 per day, delivered through residential proxies that defeated standard per-IP rate limits. Vercel deployed BotID deep analysis, blocking more than ten thousand bot requests in the first minutes. Vercel also open-sourced deepsec, a coding-agent-powered security harness scaling to 1,000+ concurrent sa…
GitHub Advisory Database Hits Record Vulnerability Volume
GitHub reported on June 29, 2026 that the Advisory Database is processing more vulnerability reports than ever before, with volume breaking records. GitHub called on the community to help manage the surge, signaling that the community-driven security model may be approaching a scalability ceiling as AI-generated code increases vulnerability surface area.
Databricks $6.9B Run-Rate and Monopoly Disruption Thesis
Databricks disclosed at SaaStr AI on June 29, 2026 a $6.9B revenue run-rate growing 80%+ year over year, with AI products alone past a $1.7B run-rate and net retention above 140%. Co-founder Arsalan Tavakoli-Shiraji stated that every software monopoly will fall in the next 24 months, providing a high-visibility data point on the pace of AI-driven disruption to incumbent developer platform vendors.
Adobe Price Increase Deferral: First Major B2B SaaS Pricing Crack Since 2022
SaaStr reported on June 24, 2026 that Adobe deferred a significant annual price increase, characterizing it as the first major crack in B2B SaaS pricing power since 2022. The prior four-year playbook of raising prices every 12–18 months and adding AI SKUs as additional charges is now under customer pressure, signaling that developer platform vendors must demonstrate clearer ROI or absorb AI costs into existing tiers.
Cloudflare Real-Time AI Spend Limits in AI Gateway
Cloudflare announced real-time spend limits in AI Gateway on June 5, 2026, integrated with Cloudflare Access for identity-driven cost governance at the user or team level. This positions Cloudflare as an enterprise AI cost governance layer at the infrastructure level, addressing the finding that nearly two-thirds of organizations lack adequate IT asset visibility to control AI costs.
示唆・見るべき論点（7件）
- 1.The bifurcation between DeepSeek V4's 17% token share at ~1% spend share and Anthropic's 65% spend share at lower token share establishes a durable two-tier AI model market: low-cost models will absorb the bulk of volume in production environments while frontier models capture economic value in quality-critical tasks. Developer platforms should architect explicit model-routing primitives that make tier-selection measurable and auditable — treating model selection as a risk management decision ra…
- 2.Vercel's Agent Stack, Atlassian's Claude Agent for Jira, and Cloudflare's Temporary Accounts collectively signal that the agentic infrastructure product category has reached naming maturity — platforms that have not yet packaged their agentic capabilities into a coherent named offering face an increasing narrative disadvantage in enterprise sales cycles where procurement teams are now evaluating agentic platform commitments explicitly.
- 3.The gap between 85% individual AI usage and 29% team-level embedding in Atlassian's survey of 12,000+ workers is not an adoption lag — it is a workflow design problem. Developer platform vendors that build team-level AI activation patterns (shared agent workflows, cross-team AI context, collaborative AI dashboards) will capture the expansion revenue that individual seat-level AI features cannot unlock, and will be better positioned as enterprise buyers scrutinize AI ROI at renewal.
- 4.Adobe's deferred price increase — the first major B2B SaaS pricing crack since 2022 per SaaStr — combined with Gartner's projection that AI coding costs will soon exceed human developer costs signals that the frictionless AI monetization era is ending. Vendors should accelerate the transition to consumption-based or outcome-based pricing models and build cost attribution dashboards into their platforms now, before enterprise buyers make granular cost visibility a procurement prerequisite.
- 5.Supabase's compounding moat — $500M in capital, official ChatGPT and Claude app status, ISO 27001 certification, 100,000 GitHub stars, and continued product velocity — illustrates the required dual credential for backend infrastructure vendors competing for agentic application workloads: developer community trust combined with enterprise compliance. Competitors lacking both dimensions will find it increasingly difficult to compete at the database layer as agentic application procurement moves in…
- 6.Cloudflare's Temporary Accounts innovation reveals a previously underappreciated bottleneck in agentic workflows: the requirement for human account setup before an agent can deploy infrastructure. Developer tooling vendors must audit all human-oriented onboarding steps that will break autonomous agent loops — the platforms that remove these friction points earliest will accumulate structural advantages as agentic workflows move from experimentation to production.
- 7.AWS's paired strategy of API compatibility (OpenAI and Anthropic client libraries) plus CloudWatch observability for bedrock-mantle eliminates both the technical switching cost and the observability gap — the two most cited reasons enterprises give for not migrating AI workloads to a new provider. This positions Bedrock as a lower-risk enterprise alternative at precisely the moment when AI budgets are scaling and CIOs are scrutinizing multi-vendor risk concentration, accelerating competitive pre…
信頼度サマリー
今週引用したソース 20 件あなたが選んだ 15 件の監視URLから検出（1つのURLから複数記事が出ることがあります）。
各ソースは信頼度レベルに応じて重み付けされています。単独ソースの主張は AI 合成時に未検証としてフラグ付けされます。
参照ソース一覧
Vercel reports weekly deployments doubled in three months, with over 30% of deployments initiated by coding agents — up 1,000% from six months ago. Claude Code accounts for 75% of agent-driven deployments. Projects deployed by coding agents are 20x more likely to call AI inference providers.
Seven months of production data from 200,000+ unique teams. Anthropic led spend at 61% in April 2026; Google led token volume at 38%. Agentic workloads carry 59% of all token volume, up from 31.6% six months prior. Teams at 10M+ requests average 35 distinct models.
DeepSeek V4 captured 17% of token share in one month while spend share stayed near 1%. Anthropic's spend share grew from 61% to 65% in May 2026. Total tokens grew +20% MoM and spend grew +43% MoM. Gemini 3.5 Flash held only 7% of Flash family tokens at month-end despite being newer.
Documents an inference theft attack on April 12, 2026 where traffic spiked to 1,300 requests per minute at 10x normal volume, projecting over $10,000/day in inference costs. Attack used residential proxies. Vercel deployed BotID deep analysis, blocking 10,000+ bot requests in the first minutes.
Vercel open-sourced deepsec, a coding-agent-powered security harness scaling to 1,000+ concurrent sandboxes on Vercel infrastructure, using Claude Opus 4.7 and GPT 5.5. Reports a false positive rate of roughly 10–20%.
Superset generates roughly 600 preview deployments per day and 1,000–1,400 deployments per week on Vercel, with 57–64% week-over-week DAU growth.
General Intelligence, an 8-person team, runs over 4,000 branches in flight with approximately 100 parallel app versions at any moment. Engineers ship an average of 10 PRs per day on a $5,000/month token budget.
Okara, a 4-person team, processes 4 billion tokens daily on Vercel, managing AI CMOs for 120,000+ businesses. New AI models are made available to users the same day they ship via AI Gateway.
Vercel launched the Agent Stack on June 17, 2026, consolidating AI SDK, AI Gateway, Sandboxes, Fluid Compute, Workflows, and Observability into a single named product for production-grade agents.
Vercel launched platform-native feature flags with server-side evaluation, automatic flag registration, and a CLI accessible to both developers and coding agents. The v0 team runs hundreds of flags at any given moment including for AI model routing.
Supabase raised a $500M Series F at a $10B pre-money valuation led by GIC on June 4, 2026. Also covers: official ChatGPT app status (May 8, 2026), ISO 27001 certification (April 22, 2026), Multigres v0.1 Alpha, npm supply chain attack guidance (May 26, 2026), and 100,000 GitHub stars.
Covers Claude Agent for Jira (June 24, 2026), Cursor integration in Jira, Bitbucket Agentic Pipelines with Claude Code, @Jira for Slack, Gartner Magic Quadrant 4x Leader for DevSecOps Platforms, AI Alert Grouping saving 839 hours in 28 days, State of Teams 2026 survey findings, and Flex commercial model.
AWS launched a redesigned Amazon Bedrock console supporting the OpenAI Responses API, OpenAI Chat Completions API, and Anthropic Messages API via the bedrock-mantle endpoint, enabling existing applications to run on Bedrock with minimal code changes.
AWS added CloudWatch metrics for the bedrock-mantle endpoint covering inference counts, input and output token totals, and client error counts at account, project, model, and project-and-model granularity.
OpenAI GPT, OpenAI GPT OSS, and NVIDIA Nemotron models on Amazon Bedrock, along with Kiro, received FedRAMP High and DoD IL-4/5 authorization in AWS GovCloud (US) on June 25, 2026.
Covers real-time AI spend limits in AI Gateway (June 5, 2026), Temporary Accounts on Cloudflare Workers (June 19, 2026), saga-style rollbacks for Cloudflare Workflows (June 29, 2026), Security Insights reaching 120+ scans per second (June 12, 2026), and Town Lake/Skipper internal AI agent (May 28, 2026).
GitHub Universe 2026 announced for October 28–29, 2026 with 'agentic era' theme. Advisory Database processing record vulnerability volumes (June 29, 2026). UNDP Ghana partnership announced June 26, 2026.
Reports Adobe deferred a significant annual price increase — characterized as the first major B2B SaaS pricing crack since 2022 (June 24, 2026). Databricks disclosed $6.9B revenue run-rate growing 80%+ YoY with AI products past $1.7B run-rate and net retention above 140% at SaaStr AI (June 29, 2026).
Reports AI coding will soon get pricier than human developers per Gartner analysis. Nearly two-thirds of organizations lack adequate IT asset visibility to control AI costs. Demand for AI-ready coders has grown almost 600% since 2021. World's top 2,000 public firms have nearly $18 trillion in untapped AI value.
Stripe Atlas data reports solo founders in the top decile generated 61x the revenue of the median solo founder in their first six months in 2025, with solo founding at an all-time high.