How many engineers will use AI coding assistants by 2028?

Gartner projects 75-90% of enterprise software engineers, up from approximately 14% in early 2024. The forecast is the most-cited enterprise-adoption number for the category.

How big is the paid AI coding assistant market?

Microsoft's FY26 Q2 SEC filing reported 4.7M paid GitHub Copilot subscribers. Cursor / Anysphere disclosed $2B ARR in early 2026 with over 1M paying users. Anthropic's Claude Code reached a $2.5B run-rate. The three-vendor sum alone defines a multi-billion-dollar paid market.

Where can I find primary-source vendor data?

Microsoft files Copilot numbers in its SEC 10-Q earnings statements. Anysphere / Cursor disclosures appear in press releases and TechCrunch coverage at funding events. Anthropic publishes Claude Code revenue in public investor communications. Always cite the primary source rather than the secondary aggregator.

Why does the market data matter to my detection-layer investment?

Detection-layer urgency tracks AI-assistant adoption. The Gartner curve sets the timeline for when every team will be shipping AI-generated code. Builders pitching detection investment to a CTO can anchor on Gartner; the audience accepts the source by default.

AI Coding Market Research

Primary-Source Data For The Detection-Layer Investment Argument

BrassCoders treats this category as the strategic-context layer. Builders defending detection-layer budget upward through their organization, founders pitching investors, or AppSec leads forecasting team growth should cite these primary-source disclosures rather than secondary commentary. The audience accepts these sources by default.

📊 Gartner — AI Code Assistant Adoption Projection 2024

Gartner, 2024 · gartner.com

BrassCoders treats Gartner's projection — 75-90% of enterprise engineers using AI code assistants by 2028, up from approximately 14% in early 2024 — as the canonical enterprise-adoption forecast. Builders pitching detection-layer investment to enterprise stakeholders should anchor on Gartner; the audience accepts the source by default.

What it's good for: enterprise-side adoption forecast. Where BrassCoders draws from it: the enterprise-adoption proof point for the detection-layer argument.

🏢 Microsoft FY26 Q2 Earnings Call — GitHub Copilot 4.7M Paid Subscribers

Microsoft Corporation, FY26 Q2 earnings call, Jan 28, 2026 · microsoft.com/investor (FY26 Q2)

BrassCoders treats this as the canonical measurement of paid Copilot adoption. The 4.7M paid-subscriber figure (up roughly 75% year over year) was disclosed verbally on Microsoft's FY26 Q2 earnings call, Jan 28, 2026 — it is NOT in the written SEC 8-K press release exhibit for that quarter, so cite the earnings-call webcast/transcript, not the 8-K, as the primary source. Builders citing Copilot adoption should note this distinction rather than imply a written regulatory filing states the number.

What it's good for: Copilot adoption data as actually disclosed (earnings call, not SEC filing text). Where BrassCoders draws from it: the lead claim in the Copilot division-of-labor post.

🏢 Cursor / Anysphere — ARR Disclosures 2025-2026

Anysphere / Cursor, public disclosures, 2025-2026 · techcrunch.com (June 2025 checkpoint)

BrassCoders treats Cursor's public ARR trajectory — passing $500M in June 2025 on the way to $2B in early 2026 — as the canonical evidence for category-2 AI coding assistant adoption (beyond the GitHub-platform-default Copilot). Builders sizing the broader AI-coding market should pair Cursor's curve with the Copilot subscriber count.

What it's good for: non-Microsoft AI coding assistant adoption sizing. Where BrassCoders draws from it: market sizing for the AI-coding category and the detection-layer messaging.

🏢 Anthropic — Claude Code Run-Rate Disclosures 2026

Anthropic public disclosures, Feb 2026, as reported by MindStudio · mindstudio.ai

BrassCoders treats Anthropic's reported $2.5B Claude Code run-rate as the third major data point in the AI coding assistant market (after Copilot and Cursor). The three-way market structure — GitHub-platform-default, IDE-native premium, terminal-native premium — is BrassCoders's working model for which AI assistants customers are using when they install BrassCoders.

What it's good for: Claude Code adoption sizing within the AI coding assistant market. Where BrassCoders draws from it: persona segmentation across AI coding assistants and the AI-tool targeting in BrassCoders messaging.

📊 The Pragmatic Engineer — AI Adoption Series 2024-2026

The Pragmatic Engineer, ongoing · newsletter.pragmaticengineer.com

BrassCoders cross-lists the Pragmatic Engineer's AI adoption series here because the publication tracks the same market BrassCoders sells into. The series is one of the few sources that survey real practitioners at scale; the methodology is publicly explained. Builders looking for ongoing market intelligence rather than a one-time snapshot should subscribe.

What it's good for: ongoing practitioner-side market intelligence. Where BrassCoders draws from it: background context across messaging and pillar writing.

📄 METR — Randomized Trial of AI Tools on Developer Productivity

METR, arXiv 2507.09089, 2025 · arxiv.org/abs/2507.09089

BrassCoders treats this as the sharpest evidence that AI-assisted speed is partly a perception. In METR's randomized controlled trial, experienced open-source developers were 19% slower with AI tools while believing they were 20% faster. Builders sizing the real cost of AI-assisted development should pair this productivity gap with the defect-rate data — the tax is unreviewed code, which a deterministic scanner is cheap enough to catch on every change.

What it's good for: the measured gap between perceived and actual AI-assisted productivity. Where BrassCoders draws from it: the counter-narrative in the speed perception-gap post.

What Does The AI-Coding Market Look Like In 2026?