Zach Aarons

The through-line

Aarons has spent his investing career on a single structural bet: that real estate and technology were two worlds that had never been properly connected, and that connecting them is a decades-long job rather than a trade. MetaProp exists, he says, because "we're looking to tether together real estate and technology which up until pretty recently those two sectors weren't really tethered together" [2]. Venture capital is one business model for doing that, and he is explicit that it is not the only one, pointing to product companies, news aggregators and events businesses as other ways to peel the same banana [2]. Behind the near-term work sits a longer mission he frames in almost civic terms: figuring out "how to ensure the long-term survivability of the built world" and "how do you make the build world more affordable," more accessible, more resilient, through public-private partnership between real estate companies, software companies and municipalities [2]. He is realistic about the pace, invoking Andy Dufresne chipping away at one small section of wall over fifty years [2].

What has changed over ten years is not the thesis but the substrate. The early framing was proptech; the current framing is broader and more industrial, covering everything that sits at the intelligence layer of real assets [3], and the dominant question has moved from software adoption to agentic AI, physical robotics and the capital discipline that follows a burst bubble [1][3]. He has also lived through the office market's difficulties and the high interest rate environment, and reads both as backdrop rather than reason to retreat, still seeing openings in AI and AR [4].

On what counts as proptech

Aarons resists the narrow definition. MetaProp backs companies "financing how the physical world gets built," plus "insuring, powering, designing, building" it, which is why the portfolio can hold a marketing and communications company like Attentive Mobile alongside hardware for water [3]. He calls the firm verticalized, but the vertical is the real asset economy rather than any one product category, and he describes MetaProp as investing at the intersection of technology and innovation and the real asset economy [1][3]. The advantage he claims is a network that runs the length of that ecosystem: real estate investors, developers, construction companies, building material manufacturers and distributors, banks and insurance companies, down to the loan originator, the insurance underwriter, the property manager, the plumber and the HVAC technician [3].

Flo is his worked example of what that network is for. It was hardware attached to a residential plumbing stack that could detect a likely leak, and MetaProp connected it across the plumbing manufacturing world, the plumbing subcontractor world and the insurance companies that ultimately became its distribution channel, ending in an exit to Moen [3]. The point is that a single deal touched four constituencies the firm already had relationships with, which is the whole argument for verticalizing.

On portfolio construction

He builds early-stage portfolios around roughly 30 core holdings and growth portfolios around 12 to 15 names, entering at pre-seed or seed and then, in his words, "saving a significant amount, about 50%, to really pile into the winners" through Series A and Series B before growth investing picks up [3]. Diversification is deliberate and multi-axis: across technological themes, across asset classes within the physical environment, and across customer base and business model [3]. Certain themes recur in every fund by design, hardware, insurance, finance, design and data [3].

What rotates fund to fund is the customer. Different buyers get up to speed at different speeds, so the timing of selling into an asset class matters as much as the technology, and he tracks the cycle plainly: data centers and digital infrastructure hot over the past few years, industrial and logistics hottest during COVID, office before that, multifamily before that [3]. Some asset types are more volatile than others in their technology adoption. Multifamily he calls a stable bet where adoption has only gone parabolic a couple of times in the history of the cycle, while anything selling into office has obviously been under pressure [3].

On discipline after growth at all costs

Aarons divides the last decade into two errors. From 2015 through 2020 and 2021, with Brazil moving roughly in lock step until 2022, the prevailing rule was growth at all costs, don't worry about spend, there will always be equity and debt capital [1]. When that capital dried up he saw an overcorrection from 2022 through 2024 into what he calls profitability at all costs, telling startups they would never raise another dime and needed to be profitable tomorrow, which for companies building advanced technology is "either impossible or very detrimental to the business" [1].

His own standard is stage- and rate-specific. A company growing six or seven hundred percent year over year and not yet EBITDA profitable is fine, and the equity capital will be there [1]. A company growing 50 to 60 percent on revenue needs to be profitable already or show a path within 12 to 18 months, because "the inconvenient truth is it's chasing outrageous growth" [1]. He has AI-era companies going from $500,000 to $10 million in revenue in a year and thinks it would be foolish to sacrifice that just because a venture capitalist demands profitability, while a mature, slower-growing company should work as hard as it can to get there [1].

On agentic AI and the shape of real estate work

His starkest claim is that "there are entire roles within the real estate ecosystem that are quickly becoming more and more agentic," heading toward no human required or very few compared with teams of dozens [1]. Leasing is where penetration is furthest along: agents can now carry a prospective tenant from browsing to asking questions to signing a lease, and he expects a robot to eventually help them move in [1]. For developers and investors, the transformative return today is the ability to evaluate hundreds of opportunities intelligently at once instead of a handful [1].

He is careful about how much has actually changed. Because the workflows are fundamentally the same, the business is not yet getting done differently. Investors are doing the same job faster, more efficiently and arguably more accurately, and "they're just helped by agents instead of being helped by humans" [1]. The redefinition of the workflow itself has not happened, and he hopes it arrives over the next ten years [1].

Building operations are the nearest thing to a genuine break. The past decade produced data, analytics and insights delivered to a building engineer who then made the call, noticing that the air conditioning is running at full blast at six on a Saturday morning with nobody in the building [1]. The next phase is an agent actually controlling the keys, with a human still monitoring and able to intervene [1]. A building engineer today can handle only one complex commercial or mixed-use asset and is perpetually frazzled; with agents, Aarons expects one engineer to manage five buildings, consuming less power, creating less pollution and making tenants more comfortable [1]. He considers a genuinely autonomous building very close, certainly in new construction [1].

On hospitality moving into the workplace

One of the shifts he has watched over MetaProp's ten years is white-glove service culture migrating from hotels into offices, where twenty years ago there were essentially no amenities and no concierges [1]. He expects agentic concierges in office buildings next, and reads the demand as being about quality of life at work: air quality, sustainability, good food options [1]. The commercial logic is that staffing concierges everywhere is a personnel-intensive business, so AI is what lets that culture diffuse beyond the top triple-A, $200-a-square-foot buildings into office at large [1].

On the hardware and materials wave that hasn't landed yet

The innovations he most wants are the ones that have not reached mass adoption: advances in material science that make building materials cheaper, more sustainable, more resilient and fundamentally easier to create and acquire, and on-site robotics that build faster, cheaper and safer [1]. He ties this directly to housing undersupply in both the US and Brazil, and notes with some frustration that these innovations have appeared in pockets without yet transforming how the industry does business [1].

He sees the next hardware wave arriving partly from an unexpected direction. Advances in defense tech, enabled by AI and by infrastructure originally built for the metaverse, VR and AR, are flowing back into ordinary construction: "we're seeing a lot of defense tech stuff coming back into just general construction," whether for utilities, data centers or affordable housing [3]. His example is a device that looks like a mini missile, burrows underground autonomously and lays telecom cabling, trenching cheaply and easily [3]. On humanoids, he thinks recording people doing household or field tasks is a smart way to gather training data, and points to an alternative energy installer planning to train its own robot line on installations its human crews already perform [3]. The more interesting question to him is when a robot stops copying the human and has its own move 37 moment, finding a better way to fold the laundry [3].

On insurance as the category being rebuilt

Insurance is where he sees the most structural opportunity, both in new types of policies and new ways to underwrite existing ones [3]. Traditional underwriting models are poorly prepared for "super extreme weather events happening everywhere all the time," with one-in-a-hundred-year events now arriving annually, which is what the Demex Group addresses by selling data and modeling into reinsurers [3]. The second wave he is underwriting is physical AI: if there are five billion artificially intelligent robots on the planet in thirty to fifty years, across factories, warehouses and homes, current general liability and operating casualty models do not cover it, which is the thesis behind Boop [3].

He is deliberately business model agnostic here. MetaProp backs insurance data and workflow automation companies such as advocate and the Demex Group, and has been bullish on MGAs and tech-enabled, now agentic, brokers and managing general agents, with a successful exit of Obi to the Baldwin group [3]. Replacing the old analog insurance company with something agentic is equally on the table [3].

On data centers and where a small fund should not compete

Aarons is candid about the limits of fund size. MetaProp cannot write cheques into neo clouds needing hundreds of millions or billions, even though he thinks many are genuinely interesting businesses [3]. What he does look at is the heating and cooling element, better cooling technologies, and unusual siting such as subterranean co-located data centers or underwater [3]. "Space data centers are cool," he says, but that opportunity in venture is well talked about, well funded, and not where MetaProp should be trying to compete [3]. The adjacent play is labour: data centers demand enormous numbers of electricians, plumbers and drywallers in rural areas with almost no skilled labour, which is why he is interested in anything connecting labour to the data center, including portfolio company Skill-it, which helps general contractors source workers [3].

On Brazil, technical debt and leapfrogging

Aarons treats Brazil as one of the clearest mispricings in the real asset economy. Only around $3 billion of data center investment has occurred there during the AI boom, despite the country having abundant clean energy, its own large oil and gas reserves, significant power infrastructure for hyperscaler data centers, a large population to build them and a great deal of vacant land [1]. He sees a path from $3 billion to $30 billion over five to ten years and describes it flatly as an opportunity for someone starting something new [1].

The deeper argument is about technical debt. "What might be perceived as a liability of a lack of software adoption in the past in Brazilian industry could also potentially be viewed as an asset today," he argues, drawing the analogy to West Africa going straight to smartphones without the janky flip phone era, so consumer applications were built mobile first [1]. The US corporate sector adopted SaaS tools rapidly over the past decade, and many of those tools will be usurped by agentic-native technologies or forced to dramatically transform their interfaces and databases; Brazil's shallower adoption curve leaves less to unwind, which could produce a leapfrogging effect in AI adoption [1].

On deal flow, the accelerator, and getting his attention

The firm's throughput is high by design. Aarons describes seeing roughly 40 deals a week, about 160 a month, against one and a half investments a month, and says MetaProp is simultaneously chasing companies and being knocked on by them, which is the healthy state for a firm [2]. Asked how a founder gets noticed, his answer is to have one of your customers reach out to him, something founders almost never do but which he wants, because it puts him in conversation with more real estate operators and is win-win [2].

The accelerator is deliberately narrow: five or six companies a year out of hundreds of applicants, run as a program rather than a live-in office arrangement, since many companies do not want their operations disrupted [2]. He kept it because he still believes proptech requires a programmatic approach to sales and distribution and to setting metrics and goals for yourself, and the main thing he wants from applicants is a clear, fundamentally business-related goal, since the program is intense and not for everybody [2].

On why range beats specialisation

Aarons attributes much of his pattern recognition to having done a bit of everything rather than mastering one function: property management, social media, acquisitions, dispositions, leasing [2]. His summary is that "I didn't get good at anything" but he can spot "problems in everything" [2]. Had he specialised, he says, he might have been excellent at leasing but unable to evaluate as many companies quickly. Having accumulated two or three hundred real estate problems over a career means there are a lot of companies already sitting in his head [2]. The same instinct for verticalized investing drew him to ENIAC Ventures around 2011 precisely because they were a mobile-focused fund at a time when that was still a differentiated category, and he watched them evolve their messaging as everything became mobile [3].

Takeaways

• Judge burn against growth rate: a company growing 600 to 700 percent a year can stay unprofitable, but one growing 50 to 60 percent needs EBITDA profitability now or a path within 12 to 18 months [1].
• Leasing is the most agentically penetrated function in real estate, with AI already carrying tenants from browsing to signed lease [1].
• Expect one building engineer to manage five buildings with agents controlling systems directly and a human monitoring, rather than agents merely reporting insights [1].
• Brazil's shallower SaaS adoption is an asset, not a liability, because it carries less technical debt into the agentic era, and its $3 billion of data center investment could plausibly become $30 billion [1].
• Build early-stage portfolios of about 30 names with roughly 50 percent reserved to concentrate into winners, and 12 to 15 names for growth [3].
• Insurance needs rebuilding twice over: for weather events that no longer behave like one-in-a-hundred-year risks, and for a world of billions of AI robots that current liability and casualty models never anticipated [3].
• A small fund should avoid neo clouds and compete instead on data center cooling, siting and the skilled labour bottleneck [3].
• The real unlock for housing costs is materials science and on-site robotics, which have appeared in pockets but have not yet changed how the industry builds [1].
• Warmest introduction to MetaProp is not a founder pitch but one of your customers emailing him [2].