Good morning,
Today’s report comes from weeks of conversations with law enforcement officials in Washington and across the country, as well as analysts and policy advisers who follow the sector closely. We wanted to understand how Axon — the company once known simply for the Taser — has transformed itself into the dominant technology platform in American policing. What follows is the story of that evolution, and the larger question of whether law enforcement is on the verge of its own “ChatGPT moment.”
On August 10, 2025, shares in Axon Enterprise reached $885.85, capping a run that multiplied the company’s market value tenfold in five years. For a business founded three decades ago as Taser International, the milestone marked more than just investor enthusiasm. It confirmed Axon’s transformation from a hardware supplier into the central software-and-data platform for US policing. Annual recurring revenue now exceeds $1.1 billion, up 34 per cent year on year, with more than $10 billion in contracted bookings extending a decade ahead. Software and services generate gross margins above 74 per cent, numbers that place Axon alongside elite enterprise SaaS firms.
SPONSORED CONTENT: Today's Report is brought to you by the Free AI:ROI Conference
THIS WEEK: Join the Free AI:ROI Conference on September 25 -- Learn How to Get Real Returns on Your AI Investments
On September 25, get the expert formula for turning AI spend into ROI at your oganization at Section's virtual AI ROI conference.
No gatekeeping, no pitching, no BS. Hear from top AI leaders like Scott Galloway, Brice Challamel, and May Habib and leave with a clear path to AI ROI for your company.
This success rests on a simple calculation. Policing has long been rich in data, but efforts to convert that data into predictive intelligence have repeatedly faltered in the courts and on the streets. The sector is primed for technology, yet bounded by constitutional limits and civil mistrust. Axon has prospered by steering clear of high-risk “predictive” products and instead embedding itself in the administrative core of policing—body-camera evidence, report writing, and digital records—areas where efficiency can be quantified and liability minimised.
The market now faces a binary question. Will policing experience a decisive AI breakthrough, akin to the productivity shock ChatGPT delivered in offices, or will tools remain confined to clerical efficiency? If the former, Axon is positioned to dominate. If the latter, its valuation is at risk.
The Long Build: Data Abundance Without Legitimacy
Policing became data-rich long before artificial intelligence became fashionable. In the 1980s, the integration of DNA testing turned forensic investigation into a probabilistic science. The 1990s brought CompStat, pioneered in New York, which institutionalised weekly metrics and geographic mapping as the basis of performance reviews. Yet CompStat also demonstrated a core weakness: the same arrests and stops that commanders were pressured to deliver as “outputs” became the inputs for future enforcement. Metrics reinforced themselves, producing circular analysis that rewarded the system for proving what it already assumed.
After the attacks of September 11th, federal homeland-security programmes funnelled billions of dollars into local law enforcement, underwriting new databases, networks, and surveillance tools. By the early 2000s, police agencies had become vast repositories of structured data. With more than 10 million arrests annually, they accumulated incident reports, field interviews, dispatch logs, and growing volumes of digital video. Georgetown University’s Center on Privacy and Technology described the result as the most comprehensive dataset maintained by government outside the intelligence community. Yet much of this information was produced by police themselves, embedding enforcement choices into the very material used for analysis.
This abundance made policing an early target for predictive analytics. Large technology companies and smaller start-ups pitched algorithms that promised to identify crime “hot spots” or high-risk individuals. The business case emphasised efficiency: with historical data as fuel, departments could allocate officers more effectively, prevent incidents, and justify shrinking budgets.
The reality was disappointing. Chicago’s “heat list,” designed to rank individuals by risk of involvement in gun violence, cost about $2.5 million to develop. Evaluations by the University of Chicago and RAND found no significant predictive value. Worse, the model disproportionately flagged residents of predominantly Black neighbourhoods, deepening mistrust and exposing the city to legal challenge.
Other jurisdictions experienced the same loop. In Los Angeles, early hot-spot pilots relied heavily on narcotics arrest data. Because those arrests had long been concentrated in a handful of neighbourhoods, the algorithms simply directed more patrols to the same streets. Far from uncovering hidden patterns, the systems validated existing assumptions. Analysts later acknowledged that these tools often produced the answers police departments expected, rather than independent assessments of risk.
Facial recognition followed a similar arc. Laboratory accuracy rates collapsed in the field, where lighting, angles, and population diversity produced misidentifications. Error rates were highest for women and minorities. Wrongful arrests triggered lawsuits and public protests, leading cities such as San Francisco and Boston to ban the technology by 2020. Yet restrictions are no longer uniform. Several states have authorised limited deployments under stricter oversight, and police unions continue to lobby for its use in serious crime investigations. The regulatory trend is volatile, not permanent, giving companies with strong compliance infrastructures new openings.
Litigation has remained the sharpest constraint. Clearview AI agreed to a $51.75 million settlement in 2022, partly paid in equity, after a wave of privacy challenges. Across the country, lawsuits linked to algorithmic bias in policing produced more than $50 million in cumulative settlements and mandated reforms. Insurers began factoring algorithmic risk into municipal liability premiums, adding hidden costs to adoption. Venture funding for predictive policing and facial recognition firms declined by about 40 per cent between 2019 and 2022, reflecting the sector’s growing risk profile.
The lesson is clear. Policing has the data to support advanced analytics, but every attempt to automate judgment or prediction runs into constitutional barriers and political mistrust. Restrictions can be tightened or relaxed, but volatility itself becomes the constant. For companies, that means compliance and liability management are as critical as technical performance. For Axon, which has built its business in lower-risk domains, it creates both a protective moat and a latent opportunity. If restrictions continue to soften and trusted frameworks emerge, it is arguably the firm best positioned to expand. If not, its growth will remain bound to administrative efficiency—and its current valuation will prove unsustainable.
Axon’s Corporate Evolution: From Weapons to Workflow
Axon’s ascent rests on a sequence of pivots that each aligned with shifts in policing and public opinion. A former police chief who adopted its body-worn cameras in the early 2010s recalled that “after every controversial shooting, we needed something that signalled accountability — the cameras gave us that.” For Axon, cameras were not just hardware but a Trojan horse for software: once deployed, departments needed secure storage, disclosure tools and chain-of-custody systems. Evidence.com became the indispensable subscription service.
Investors say this was the real transformation. “It stopped being an equipment vendor and started looking like a SaaS utility for public safety,” said one technology analyst who covers both enterprise software and defence contractors. That view is borne out in the numbers: software and services revenue rose from $77mn in 2020 to $312mn in 2024, while annual recurring revenue reached $1.1bn in early 2025. A net revenue retention rate of 123 per cent — more common in high-growth enterprise SaaS than in government IT — means departments keep adding modules once they are on the platform.
The company’s approach to AI has been similarly pragmatic. Rather than building predictive policing tools that invite equal-protection lawsuits, Axon launched Draft One, which turns body-camera footage into draft reports. A California officer who piloted the system said it “cuts hours off a shift — the report practically writes itself.” Municipal managers find the calculation straightforward: compare subscription costs of $200–$500 per user per month against overtime and staffing. For Axon, the critical point is that the product accelerates paperwork without dictating enforcement.
Market Dynamics: Compliance as Competitive Advantage
The global market for police AI is projected to expand from $3.4bn in 2024 to $157bn by 2034, but growth is shaped less by technical progress than by legal durability. A venture investor who exited two facial-recognition start-ups described the change bluntly: “Five years ago, pitches were about algorithmic power. Today, the first half-hour is your general counsel explaining liability cover.”
This has narrowed the competitive field. IBM, Microsoft and Amazon retreated from police-facing applications in 2020, wary of political blowback. Palantir remains a force in federal contracting, but municipal buyers are cautious. “City councils don’t want to wake up in the papers for using a system they can’t defend in court,” one procurement adviser said. Smaller challengers like Flock Safety and ZeroEyes specialise in licence-plate readers and school security, but lack Axon’s scale. As a former DOJ official noted, “Niche vendors can’t produce the compliance documentation that chiefs now demand. That’s Axon’s moat.”
International markets show the same dynamic. Chinese groups Hikvision and Dahua command more than 40 per cent of the surveillance market globally but are excluded from US procurement under the NDAA and FCC restrictions. In Europe, the new AI Act requires that police systems be registered as “high-risk” and subjected to rights impact assessments. A Brussels-based consultant put it plainly: “Selling here isn’t about translation, it’s about re-engineering your product for transparency.” That creates high entry costs but premium pricing opportunities for compliant vendors.
For Axon, the mix is protective at home and costly abroad. Its domestic market is shielded from low-cost rivals, and federal grants provide steady funding. Overseas expansion requires duplicative compliance infrastructure, eroding margins. Still, analysts argue that if any US company can export a “compliance-grade” policing platform, it is Axon.
Current Deployments: Efficiency, Not Prediction
The reality of AI in policing is narrower than its marketing once suggested. The bold vision of systems that forecast crime has given way to tools that accelerate paperwork and manage evidence. Draft One, Axon’s automated report writer, is now its most visible AI product. Officers testing it say the appeal is pragmatic. One sergeant described it as “a second set of hands that turns video into words before I’ve even sat down.” Departments report time savings of six to twelve hours per officer each week — figures that translate directly into reduced overtime.
Evidence management is another growth area. Software can transcribe audio from interviews, tag objects in body-camera footage, and highlight inconsistencies across case files. Prosecutors say this reduces bottlenecks in disclosure, while investigators avoid sifting manually through hours of video. These tools address chronic staffing shortages in administration rather than the frontline.
Predictive systems have not disappeared, but they have shifted from people to places. Instead of producing “heat lists” of individuals, departments are experimenting with tools that forecast busy times or map likely hot spots around large events. A criminologist noted that “moving from individuals to places lowers the equal-protection risk, but it also lowers the claims of transformation.” These systems help allocate patrols, but they do not redefine policing.
Even incremental deployments create legal and cultural tensions. Courts are beginning to scrutinise AI-assisted reports. Scholars warn that if algorithms shape narratives, questions arise about authorship and reliability. Police unions have negotiated safeguards that require human oversight and prevent automated outputs from being the sole basis of discipline. Within departments, generational divides are visible: younger officers tend to embrace automation, while veterans are sceptical of anything that challenges their judgment.
Investment Case: Growth With Structural Risk
Axon’s financial profile places it among the elite of SaaS providers. With ARR at $1.1bn, net revenue retention at 123 per cent, software margins above 74 per cent, and more than $10bn in contracted bookings, revenue visibility is unusually high for a government-facing company. Federal programmes such as the COPS Technology and Equipment Program provide steady demand even when municipal budgets tighten. Domestic markets are shielded from Chinese competition.
Yet investors remain cautious. One New York–based fund manager said, “This looks like Salesforce economics, but with the politics of policing layered on top.” Roughly two-thirds of Axon’s revenue ultimately depends on public budgets. A single Supreme Court decision could redefine the permissible scope of AI in evidence handling or patrol planning. City-level bans can spread quickly by precedent. International expansion requires country-specific compliance infrastructure, raising costs and slowing scale.
Analysts see the market as binary. A London-based technology strategist put it this way: “If AI in policing gets its ChatGPT moment, Axon is the platform of record and the stock doubles again. If it doesn’t, it becomes a sticky IT vendor with a SaaS multiple that can’t hold.” The risk is that investors are already pricing in transformation that may never arrive.
Axon has outpaced its rivals by anticipating the limits of constitutional law and shaping its products around them. From Tasers to body-worn cameras to evidence management, each pivot has embedded the company deeper into police workflows while avoiding the legal traps that sank competitors. Its AI products continue this pattern: selling time savings rather than predictions, efficiency rather than enforcement.
But the valuation now rests on whether policing will ever experience a genuine AI inflection. In offices, ChatGPT unlocked productivity with little legal friction. In policing, rights are the friction. If the courts and communities permit AI to move beyond paperwork into trusted decision support, Axon is positioned to dominate a market projected at $157bn by 2034. If not, the company remains valuable but less exceptional, and its multiple will contract sharply.
The lesson is clear. The future of police AI will be written as much in courtrooms and city councils as in code. Axon is better placed than any competitor to capture the upside if restrictions loosen. But if the breakthrough never arrives, its investors will discover how unforgiving public-sector technology markets can be.