Magic.dev
Magic.dev is a startup building an AI coding agent with long context, still unreleased after four years.
On the record
Checkpoint: Deadline for Magic to ship a GA product with published benchmarks and named customers before pressure to sell or shut down builds further
The call
Magic.dev has raised half a billion dollars on a long-context bet that frontier labs now ship natively, while rivals Cognition and Cursor have already shipped and scaled past it.
Magic.dev is trying to build artificial intelligence software that writes code for programmers. The problem is that other companies like OpenAI, Google, and Anthropic have already shipped working AI coding tools that millions of people use, and these giant companies can keep improving their tools faster than Magic can. Magic's only claim is that it has special long-context technology, but that is still just an idea on paper, not something proven to work better than what the giants are already selling. The one thing that could change everything is if Magic releases a finished product that real companies want to buy and pay for in the next year or so, with proof that it actually works better than Cursor, Devin, or GitHub's Copilot.
Why
01
AI pressure: Google Gemini, OpenAI Codex, Anthropic Claude Code. The frontier labs ship long-context and agentic coding every quarter, eating the exact gap Magic was built to fill before Magic ships anything. Could AI do the core job? Yes, likely.
02
Bigger players: Cognition (Devin, ~$26B) and Cursor (~$50B). Both rivals have shipped, scaled, and monetized while Magic remains pre-revenue, turning Magic into an afterthought in its own category. Could a giant copy it? Yes, easily.
03
What protects it. None found. No shipped product, no customers, no proprietary data, and no lock-in of any kind.
What would change this call
Magic ships a GA product with published third-party benchmarks within 12 months. Would show Magic can compete commercially instead of remaining research-only.
Primary press release confirming the rumored $550M July 2025 Series D. Would show Magic secured funding runway to keep operating.
Magic announces named enterprise customers or case studies. Would show Magic has found a commercial foothold against Cursor and Cognition.
Primary press release confirming the rumored $550M July 2025 Series D. Would show Magic secured funding runway to keep operating.
Magic announces named enterprise customers or case studies. Would show Magic has found a commercial foothold against Cursor and Cognition.
Key numbers
- $465-515M raised through Series C
- No GA product or revenue confirmed
- ~47 employees vs Cursor's $50B scale
Named risks
- No shipped GA product or revenue after 4 years and ~$465M raised as of August 2024.
- Core long-context differentiation unpublished/unverified while frontier labs ship comparable context windows natively.
- Rumored $550M July 2025 Series D remains unconfirmed by any primary press release as of mid-2026.
- Valuation gap vs peers widening fast: Magic ~$1.4-1.5B (Aug 2024) vs Cognition ~$26B, Cursor ~$50B (mid-2026).
Rivals, as read in July 2026
| Rival | Direction | Evidence |
|---|---|---|
| Cognition AI (Devin/Windsurf) | losing to | Reached ~$26B valuation and shipped Windsurf 2.0, April 2026, vs Magic's no GA product |
| Cursor (Anysphere) | losing to | ~$50B valuation and ~$100M revenue by mid-2026 vs Magic's zero disclosed revenue |
| Google (Gemini) | losing to | Ships native 10M-token context windows, 2024, undercutting Magic's core long-context claim |
| OpenAI, Anthropic, Microsoft/GitHub | losing to | All shipping agentic coding at platform scale by 2026, e.g. Copilot Studio 160,000+ orgs, May 2026 |
| Codeium/Windsurf | contesting | Absorbed into Google and Cognition in 2025 after reaching ~$2.85B valuation, Feb 2025 |
Questions people ask
- Is Magic.dev profitable in 2026?
- No. As of the Series C close on August 29, 2024, TechCrunch and Reuters reporting stated Magic 'has no revenue and no product for sale.' As of mid-2026, Magic still has not announced a general-availability commercial product, so there is no confirmed path to profitability.
- Has Magic.dev shipped a product yet?
- No. As of June 2026, Magic has not shipped a general-availability product comparable to GitHub Copilot, Cursor, or Claude Code. Its public releases remain research previews: LTM-1 (June 2023, 5M-token context) and LTM-2-mini (August 2024, 100M-token research preview), neither commercialized.
- Who is Magic.dev's biggest competitor?
- Cognition AI (Devin), reportedly valued around $26B by May 2026 after acquiring Windsurf's assets for $250M and shipping Windsurf 2.0 in April 2026, and Cursor (Anysphere), reportedly valued around $50B in mid-2026 with about $100M in revenue, both dwarf Magic's last confirmed ~$1.4-1.5B valuation from August 2024.
- How much money has Magic.dev raised?
- Magic raised roughly $465M through its Series C, per TechCrunch: $5.1M seed (March 2022), $23M Series A (Feb 2023), $117M Series B at about $500M valuation (Feb 2024), and $320M Series C led by Eric Schmidt (August 29, 2024). A later $550M figure tied to a July 2025 Series D remains unconfirmed by any primary press release.
- Could AI replace Magic.dev's business?
- Magic's entire product is itself a long-context AI model, making it directly exposed to commoditization: Google's Gemini already ships up to 10M-token context windows natively (per August 2024 comparisons), while Magic's own 100M-token LTM-2-mini remains an unreleased research preview as of mid-2026.
- Who owns Magic.dev?
- Magic AI, Inc. is privately held and venture-backed, founded March 2022 by Eric Steinberger (CEO) and Sebastian De Ro. Investors include CapitalG, Sequoia, Eric Schmidt, Nat Friedman and Daniel Gross (NFDG), Elad Gil, Atlassian, and Jane Street, per its August 29, 2024 Series C announcement.
Timeline
- 2026-06-27Still no GA commercial product announced, per aiwiki summary of company status
- 2025-07-24PitchBook lists unconfirmed Series D dated, $550M figure, no primary press release found
- 2024-08-29Series C closes at $320M led by Eric Schmidt; Google Cloud/NVIDIA supercomputer partnership announced
- 2024-08-28LTM-2-mini unveiled: 100M-token context research preview, not a shipped product
- 2024-07-02Reuters reports Magic in talks to raise at $1.5B valuation
- 2024-02-01Series B closes at $117M, valuing Magic around $500M
- 2023-06-06LTM-1 announced: 5-million-token context model for whole-repo code understanding
- 2023-02-06Series A closes at $23M, total raised reaches $28M
- 2022-03Magic AI, Inc. founded by Eric Steinberger and Sebastian De Ro
Revision history
20 July 2026
First assessed. Little protection left. The text above is fixed as of this date and is not edited afterwards.
In the same band
Sources
- Magic (AI software engineer) (2026-06-27)
- Generative AI coding startup Magic lands $320M investment from Eric Schmidt, Atlassian and others (2024-08-29)
- Exclusive-AI coding startup Magic seeks $1.5-billion valuation (2024-07-02)
- Magic 2026 Company Profile: Valuation, Funding & Investors
- Magic AI Revenue, Valuation & Funding History (2024-08-05)
- Magic AI, Inc. Funding & Investors (2026-07-14)
- Magic (magic.dev) Teams
- Introducing LTM-1 (2023-06-06)
- 100M Token Context Windows, Magic (2024-08-28)
- Against Copilot-Style AI, Enterprise DNA (2026-06-13)
- AI-coding startup Codeium in talks to raise at an almost 3B valuation (2025-02-19)
- The 2026 AI Agent Vendor Landscape at a Glance (2026-05-30)