Vannevar Labs
Last reviewed
Jun 3, 2026
Sources
12 citations
Review status
Source-backed
Revision
v1 · 2,049 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 3, 2026
Sources
12 citations
Review status
Source-backed
Revision
v1 · 2,049 words
Add missing citations, update stale details, or suggest a clearer explanation.
Vannevar Labs is an American defense technology company that builds artificial intelligence and machine learning software for national security and intelligence work, with a focus on foreign-language open-source intelligence (OSINT), maritime domain awareness, non-traditional intelligence collection, and counter-disinformation. Founded in 2019 and based in Palo Alto, California, the company sells its products to the U.S. Department of Defense and the intelligence community, and it is often grouped with Anduril and Palantir as part of the wave of venture-backed firms trying to modernize how the Pentagon buys and uses software. Its flagship product, Decrypt, ingests foreign news, social media, and other publicly available data, then uses natural language processing to translate and analyze it for military and policy decision-makers.
The company is named after Vannevar Bush, the engineer who ran the U.S. Office of Scientific Research and Development during World War II, coordinated the Manhattan Project, and later wrote the influential essay "As We May Think." The name signals the founders' stated ambition: to rebuild the kind of close partnership between American scientists, technologists, and the military that existed in the mid-twentieth century.
Vannevar Labs was founded in April 2019 by Brett Granberg and Nini Hamrick, who met while studying at the Stanford Graduate School of Business. A third co-founder, Danny Goodman, served as chief technology officer in the early years before leaving around 2021 to start the autonomous aircraft company Swarm Aero. [1][2]
Granberg, the chief executive, had worked on defense strategy at McKinsey and then as an investor at In-Q-Tel, the venture arm of the CIA and the broader U.S. intelligence community. Hamrick, the president, spent years inside that community herself, including time at the Defense Intelligence Agency and the Office of the Director of National Intelligence. (She has also been referred to in earlier records under the name Nini Moorhead.) Their shared diagnosis was that the most capable commercial AI was being built in Silicon Valley while the agencies responsible for national security were stuck with legacy systems and slow procurement. One detail the founders return to often: parts of the U.S. nuclear command system still relied on eight-inch floppy disks until 2019. [1][3]
The pitch was that modern conflict increasingly plays out through software, information, and data rather than only through tanks and aircraft, and that the United States was falling behind rivals such as China in applying machine learning to that contest. Vannevar set out to close the gap by writing intelligence software that frontline analysts would actually want to use. [3]
Vannevar's core idea is to turn the firehose of open-source and publicly available information into something an intelligence officer can act on quickly. According to reporting by MIT Technology Review, the company's systems collect terabytes of data each day, spanning roughly 80 languages from about 180 countries. The inputs include foreign-language news, social media accounts, data leaked from foreign networks, non-classified reporting from human sources, and physical sensor feeds such as radio signals used to detect illicit shipping. [4]
On top of that data, the company runs a mix of large language models, including systems from OpenAI and Microsoft alongside its own proprietary models. The software translates foreign material, flags potential threats, tracks specific topics such as fentanyl supply chains or rare-earth minerals, and runs sentiment analysis through a chat-style interface that analysts can query in plain English. In one example described publicly, a Marine officer used the system to monitor foreign news mentions of a U.S. exercise and gauge how it was being received, while another used it to draft daily and weekly intelligence summaries. [4]
That approach has drawn scrutiny from AI safety researchers. Heidy Khlaaf, a machine-learning safety expert, has cautioned that large language models are unreliable in safety-critical settings and that sentiment analysis is inherently subjective, arguing that a "human in the loop" cannot realistically catch every error when a model is summarizing thousands of data points. Chris Mouton of the RAND Corporation has similarly noted that current AI still struggles to reliably detect subtle propaganda. Vannevar's leaders frame the technology as decision support rather than autonomous judgment: chief technology officer Scott Philips has described the goal as helping the military "collect data, make sense of that data, and help the US make good decisions." [4]
Vannevar started with a single product and expanded into several mission areas. The table below lists the products that have been publicly described; the company discloses limited detail about most of them given the customer base.
| Product | Focus area |
|---|---|
| Decrypt | Foreign-language OSINT: search, translation, and analysis of news, social media, and other public data |
| Serra | Non-traditional intelligence collection and analysis |
| Revere | Maritime domain awareness |
| Foreshadow | Maritime domain awareness |
| Overwatch | Foreign disinformation and influence detection |
| Local | Foreign disinformation and influence detection |
| Telescope | Strategic competition and emerging-threat analysis |
| TIE | Narrative analysis (beta) |
| Curator | Translation tooling (beta) |
Decrypt is the flagship. Launched in January 2021, it began as a tool built around Arabic optical character recognition before pivoting toward the languages of state competitors such as Mandarin and Russian. The company describes it as letting military and foreign-policy users understand, target, and respond to foreign actors in ways that were not previously possible. [2][5] Vannevar has said its more recent work pushes toward what it calls agentic systems, with a stated 2026 goal of unifying its AI capabilities into a single platform tailored to specific missions. [6]
Vannevar's first deployment was a small pilot worth roughly $25,000 covering four users over three months. It says that pilot converted into a contract worth about $1.3 million within a few months, the kind of land-and-expand path the company has held up as a model for other defense startups. [2] By early 2023, Decrypt had been fielded to personnel at more than 15 U.S. military bases worldwide and had generated about $25 million in sales. [5]
Reported and confirmed customers and commands include U.S. Indo-Pacific Command (INDOPACOM), U.S. Special Operations Command (SOCOM), Air Force Special Operations Command (AFSOC), the National Air and Space Intelligence Center, the U.S. Air Force, the U.S. Marine Corps, and the U.S. Navy. The 15th Marine Expeditionary Unit, a force of about 2,500 Marines, has used the software aboard Pacific ships operating near South Korea, the Philippines, India, and Indonesia. Vannevar's tools have also been fielded in large joint exercises, including Balikatan 2025 and Talisman Sabre 2025. [2][4]
The company's most significant contract to date came on November 18, 2024, when the Defense Innovation Unit awarded Vannevar a production-scale Other Transaction agreement with a ceiling value of $99 million and $16 million obligated at award. The deal, tied to DIU's Real-Time Information and Effects program, moved Vannevar's technology from prototype into production and is meant to let the company scale its software across the Department of Defense. The work emphasizes operationalizing commercially and publicly available information, including foreign-language content, to deliver real-time analysis, with particular relevance to distributed maritime operations in the Indo-Pacific. [4][7][8] In June 2025, Vannevar partnered with the mobile carrier and connectivity firm Cape to bring real-time radio-frequency sensing closer to the tactical edge. [6]
Vannevar has raised money across a seed round and two priced venture rounds, with investors that span generalist venture firms and defense-focused funds.
| Round | Date | Amount | Lead / notable investors |
|---|---|---|---|
| Seed | Late 2019 | $4.5 million | Costanoa Ventures, General Catalyst, Point72 Ventures |
| Series A | May 2021 | $12 million | Costanoa Ventures and Point72 (lead); General Catalyst |
| Series B | January 18, 2023 | $75 million | Felicis (lead); DFJ Growth, Aloft VC (new); General Catalyst, Point72 Ventures, Costanoa Ventures, Shield Capital (existing) |
The Series A brought total funding to roughly $16.5 million at the time. [9] The $75 million Series B was led by Felicis and valued the company at about $575 million, bringing total capital raised to more than $90 million. Felicis founder Aydin Senkut said Vannevar had "already demonstrated a strong ability to get innovative tech into the hands of the defense community." Granberg said the round would let the company "much more aggressively invest in R&D for the products we're fielding with the government." [5][8][10]
By 2025, Vannevar's growth had pushed its reported worth far above the Series B mark. In company materials and a September 2025 founder discussion, Granberg referred to Vannevar as a roughly $1.5 billion company, a figure that reflects later valuation rather than a publicly detailed new priced round. [11] Public trackers put cumulative funding in the range of $87 million to $91.5 million as of early 2026. Independent figures should be treated with some caution, since the company does not publish a full financial breakdown.
Brett Granberg is chief executive and Nini Hamrick is president. Scott Philips, who joined as chief technology officer, previously held a leadership role at Systems and Technology Research and worked at Palantir and MIT Lincoln Laboratory; he holds a PhD in electrical engineering from the University of Washington. [4][12]
The company has deliberately recruited from both the national security world and big technology firms. Its early hires included Scott Sanders, who came from Anduril to lead business development, and its board added John Doyle, a former Palantir executive who had run that company's national security business. By late 2024 Vannevar employed more than 150 people, including military and intelligence veterans alongside alumni of Amazon, Apple, and established defense contractors; that headcount grew past 200 by the end of 2025. [2][9]
Vannevar has shared selective revenue figures that show rapid growth from a small base. The company reported about $3.5 million in revenue in 2021, roughly $49 million in annual contract value in 2023, and on the order of $80 million in revenue in 2024. It has said it reached profitability within about three years of founding, which it contrasts with the longer path that Palantir took to become profitable. As with the valuation, these numbers come largely from the company and secondary trackers rather than audited public filings. [2]
Vannevar sits in a specific niche within the new defense technology sector. Where Palantir built broad data-integration platforms for governments and Anduril concentrates on autonomous hardware and weapons systems, Vannevar focuses on AI-native, multilingual open-source and signals intelligence, an area both larger competitors touch but do not fully own. [3] The company is part of a broader argument, made loudly by founders across the sector, that the Pentagon should buy commercial software the way it buys other modern technology, on faster timelines and through flexible contracting vehicles such as Other Transaction agreements. Whether software-first intelligence tools live up to that promise, especially as they lean more heavily on large language models in high-stakes settings, is still being tested in the field. [4][6]