Microsoft / GitHub
Sorbet currently classifies Microsoft / GitHub as the leader in its Source control bucket. The detailed selection note is not yet available.
No recent material read
Based on 0 material signals in the last 30 days: 0 positive, 0 negative.
Evidence summary, not a buy / sell rating
Latest approved decision: hold · 2026-08-29
The role and pillars explain what job this holding has inside the System-of-Record basket—not whether it suits an individual investor.
GitHub is the world's dominant source-control substrate, hosting the commit histories, pull-request state machines, branch-protection rules, and code-review trails that constitute the most durable, machine-readable record of how software systems change over time. Every pull request is a textbook P2-adjacent state machine — opened, reviewed, approved, merged, or closed — with typed ownership (P3), explicit API verbs such as create_pull_request, merge, and request_review (P4), and a queryable audit trail stretching back through every diff, comment, and CI status (P5); the repository itself is a P1 durable record in the purest sense, structured around commits rather than unstructured prose. Microsoft's ownership of GitHub means it sits at the origin point of every AI coding agent's context window: Copilot Workspace, third-party coding agents, and autonomous CI pipelines all resolve against the same repo records that GitHub already owns and versions. As the bucket leader, Microsoft anchors the source-control allocation — the 40% leader tilt reflects that no realistic challenger can replicate the network effects of 100-million-developer commit history, the breadth of third-party integrations, or the depth of Actions-based workflow state that agents increasingly read and write. Holding MSFT here also fulfils the thesis's explicit hyperscaler-hedge role: if bundling pressure compresses standalone SoR players, Microsoft is the bundler, not the bundled.
The bucket's primary paper position: the company Sorbet currently treats as the clearest expression of this workflow category.
The thesis properties this holding is selected to test. These labels are portfolio classifications, not standalone proof.
Durable records
Typed ownership
Structural verbs
Queryable history
The thesis for Microsoft / GitHub breaks structurally — not merely underperforms — under three concrete conditions. First, if a foundation-model provider ships a fully self-contained code-state substrate (commit graph, branch ownership, PR state machine, code-review audit trail) that customers demonstrably re-platform onto at scale, GitHub's role as the durable record of system change is displaced rather than complemented; the signal to watch is enterprise migration rate away from GitHub repositories, not Copilot attach-rate softness. Second, if GitHub's agent-facing monetisation — Actions consumption, Copilot Workspace seat growth, MCP-accessible API call volume — decelerates in absolute terms for two or more consecutive quarters while Microsoft's own Azure AI foundry continues to grow, that would indicate the source-control SoR is being absorbed as a loss-leader rather than monetised as infrastructure, eroding the P4 structural-verb and P5 queryable-history pillars that make GitHub a control plane rather than a commodity host. Third, the open-source invalidator is most pointed here: Gitea, Forgejo, and self-hosted GitLab already provide credible clones of the repository record model, and a measurable compression in GitHub's per-seat or per-compute gross margin — particularly in the SMB and developer-tool segments — would signal that pricing power over the source-control SoR is eroding faster than agent-SKU attach can compensate. Underperformance of MSFT shares relative to peers is explicitly not the trigger; the falsifier is a structural break in GitHub's role as the canonical commit-history substrate that agents read and write.
Daily closes are shown for context. Reference levels are deterministic outputs of Sorbet's published ATR method, not analyst targets or advice.
2025-07-28 → 2026-08-31
$507.29
Short-term levels use 14d ATR and the 200d average. The 3y projection compounds the current revenue-growth rate with no multiple expansion. Missing inputs stay blank.
Revenue growth + FCF yield, not the conventional FCF-margin formulation. Visible for judgment, but not wired into selection.
Price is above both the 50d and 200d averages, with the faster average leading.
| Quarter | Revenue YoY | Gross margin | FCF margin | Operating margin |
|---|---|---|---|---|
| 2026-03-31 | 18.55% | 67.63% | 19.07% | 46.28% |
| 2025-12-31 | 16.84% | 68.04% | 7.24% | 47.02% |
| 2025-09-30 | 18.49% | 69.05% | 33.04% | 48.85% |
| 2025-06-30 | 18.30% | 68.58% | 33.45% | 44.84% |
| 2025-03-31 | 13.44% | 68.72% | 28.97% | 45.64% |
| 2024-12-31 | 12.27% | 68.69% | 9.32% | 44.21% |
Microsoft's path forward in the source-control bucket runs through GitHub's ability to deepen its role as the authoritative agent substrate — not merely a hosting layer but the durable control plane through which autonomous coding workflows read history, gate merges, and close the loop on work items. The already-live Copilot Workspace and GitHub Models integrations show the direction: agents that treat the PR as a state machine, moving it from draft through review to merge via named, auditable transitions (P4, P5), with ownership tracked at every step in a typed assignee field (P3). What remains unproven is whether Microsoft can convert that architectural advantage into durable consumption revenue at scale — the agent SKU must demonstrate that richer agentic usage correlates with measurable seat or Actions-minute expansion, not just feature attachment. The hyperscaler hedge the thesis explicitly acknowledges is already embedded here: if bundled Azure + GitHub + Copilot pricing crowds out standalone challengers, Microsoft captures the compounding rather than merely enabling it, but that same bundling dynamic could obscure per-unit economics and make it harder to isolate the SoR substrate's contribution to growth.
No material signals for MSFT in this period.
positive signal but thin coverage (only P4) — hold for confirmation.
1 negative pillar (P4) — MSFT is at or below target — no trim
1 negative pillar (P4) — MSFT is at or below target — no trim
Missing data. No owner-reviewed NRR disclosure is available. Unreviewed AI extraction is never shown publicly.