Loading today’s stories
Invest2 distinct publishers3 min readPublished
Tripling off a $100m base makes enterprise context a budget line the model vendors now have to attack directly. But part of that $300m is consumption revenue. Glean's own sales pitch is designed to shrink it.
The Investor · Invest desk
Compiled by The InvestorSomething wrong?How this is made
The cadence is what makes this legible: $200m of new run rate spread across 15 months is roughly $13.3m added every month without a visible pause [1], and a tripling over that window annualises to about 2.4 times a year [2], which is the figure a competing knowledge-layer pitch now has to be measured against rather than the softer claim of category leadership.
Price has to be weighed against value here. The last mark was $7.2bn on a $150m Series F in June [12], so the company is carried at about 24 times its current run rate [3], and that mark is now eleven months stale against a top line that has moved a long way since.
That benchmark is worth a second look. Glean says it was preferred roughly 2.5 times as often as off-the-shelf MCP tools, that those tools consumed 30% more tokens on average, and that its win rate rose from 66% on simple tasks to 73% on multi-step ones [5]. Thirty percent more on the other side of the ledger is about 23% fewer on Glean's [4], and the 2.5x preference is not an independent data point so much as the same wins restated as odds, since 70 against 30 is about 2.3 to one [5].
Pricing is the term that deserves close attention here. Arvind Jain describes both a consumption model and a hybrid one, where a fixed monthly fee for active users sits beside separate usage fees for model consumption [14], while the headline benefit he sells is that customer AI bills fall because the AI performs fewer operations [16]. Sell that well enough and one of your own line items contracts: the per-active-user fee is the durable part, and the consumption fee is the part Glean is actively teaching its buyers to reduce. TechCrunch made the adjacent point that a consumption model has no strictly recurring component, so some portion of the $300m reads better as an annualised run rate than as ARR [15]; Glean had not responded to its request for comment at publication [17].
There is an obvious counter-thesis: Google, Microsoft, OpenAI, Anthropic, Salesforce and Atlassian are all building versions of this [11], and each can fold a context graph into seats a customer already pays for, which is exactly how a separate budget line reverts to a feature. Jain's answer is that the first four or five years were competition-free [10] and that the moat is connectivity into environments others decline to touch [8]. Note where the spending went: more than 250 features in a year [6], plus a Model Hub and an MCP Gateway that let customers swap models, clouds and frameworks [7]. That is a company buying plumbing and governance instead of a model.
This is probably wrong in one specific way, but I read the 45% weekly-to-monthly active ratio, said to be more than twice the SaaS benchmark [4], and the more than 85% of customers spanning five or more departments [3], as firmer evidence of a real budget line than the revenue number is, because horizontal usage is what survives the procurement review that consolidates AI spend. What would falsify it: a run rate that flattens at constant customer count, which is the consumption tell.
Ranked by verification strength, evidence, and original report placement.
Glean announced on May 28, 2026 that it had reached $300 million in annual recurring revenue, 15 months after reaching $100 million ARR.
TechCrunch describes the $300 million figure as a three-fold increase from the $100 million milestone reached 15 months earlier at the seven-year-old startup.
Jain says connecting AI to Glean results in the AI consuming far fewer tokens because it performs fewer operations, and that customers like that Glean can reduce their AI bill significantly.
Glean's platform includes a Model Hub and MCP Gateway, giving customers choice across models, clouds, frameworks, repositories and front-end experiences.
CEO Arvind Jain told TechCrunch: "The first four or five years of our existence, we had no competition."
TechCrunch lists Google, Microsoft, OpenAI, Anthropic, Salesforce and Atlassian as tech heavyweights building Glean-like enterprise search tools.
Distinct publishers with included, body-backed reporting in this cluster.
1 article · September 2, 2026
1 article · September 2, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
invest
Box's billings ran eight points ahead of revenue. The AI bill was 20 basis points.1 distinct publisher
build
Agent Plugins 1.0.0 standardises file paths. Anthropic still owns the behaviour.1 distinct publisher
build
Cohere prices document parsing at 15 percent of the hyperscaler rate1 distinct publisher
build
Grok 4.6 lands in Copilot two days after launch, and the model picker becomes a procurement problem1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One company's numbers, one reporter's caveat
Only the headline figure appears in two places, and even TechCrunch has it from Glean. Everything textured — the 45% weekly-active ratio, the 2.5x preference, the 250 features, the Fortune 500 doubling — appears once, in the company's release, with no denominators, no named benchmark and no evaluator. TechCrunch's contribution is a re-labelling rather than a check: it tells you the number is partly a run rate, not whether the run rate is real.
Real spend, unaudited breadth
Money is the sturdiest evidence in this story: tripling revenue in five quarters means invoices someone approved, and TechCrunch puts names to a few of them in Databricks, Reddit, Pinterest and Samsung. Below the revenue line the picture thins fast — the departmental spread, the Fortune 500 growth and the 28-country footprint are Glean counting itself, with no base to compare against and no word on whether early accounts renewed.
Overstated at the edges, not the middle
The headline survives contact better than the language wrapped around it. 'Unrivaled', 'unmatched' and a moat the company defines for itself do no work, and a 2.5x-preferred benchmark with no published method should not move anyone's view. The real stretch is a single word: Glean calls the whole $300 million recurring while its own CEO describes pay-per-use pricing to TechCrunch — and the pitch attached to that pricing, that Glean shrinks your token bill, is an argument for the consumption slice getting smaller.
Written to be quoted
This is a milestone release from a company that priced at $7.2 billion last June, and it ends with a registration link for its own conference — the audience is the next round and the next renewal. Two things pull the other way and are worth crediting: Glean is selling cost reduction, which eats into the usage-based part of its own top line, and it left TechCrunch's follow-up unanswered rather than managing the harder question about what 'recurring' covers.
Trust the trajectory, hold the detail loosely
Two accounts, one of them the subject, dated consistently and agreeing on the arithmetic: the direction and rough size of this business are safe to repeat. Product quality, engagement and customer mix rest on a single interested count, and the composition of the $300 million — the one thing that would settle how to read it — remains unreported by either publisher.