Atmos was a San Francisco-based homebuilding marketplace that operated from approximately 2018 until March 2025. The company positioned itself as an end-to-end platform for custom home design and construction, connecting buyers in Sun Be…
Atmos was a San Francisco-based homebuilding marketplace that operated from approximately 2018 until March 2025.[1][2] The company positioned itself as an end-to-end platform for custom home design and construction, connecting buyers in Sun Belt and Southeast markets with vetted builders, handling design approvals through 3D technology, and coordinating financing — all under one digital roof.[3] It graduated from Y Combinator's Summer 2020 batch and raised approximately $20 million across three rounds, including a $12.5 million Series A led by Khosla Ventures.[4]
Atmos failed for two compounding reasons. Its technology never automated away the human designers and project coordinators at the core of its service, leaving it structurally exposed as an operationally heavy business. When the Federal Reserve's 2022–2023 rate hiking cycle made custom home financing unaffordable for its target customers, the pipeline of months-long design engagements collapsed without converting to revenue.
The company shut down around March 2025 after completing a lifetime total of roughly 50 homes and designing over $200 million in projects that never broke ground.[5] CEO Nicholas Donahue subsequently launched Drafted, an AI-first floor plan generator with no designers on staff — the product he described as "the thing we always wanted to build."[6]
Nicholas Donahue grew up inside the homebuilding industry. His father built homes for major developers; his mother sold homes to big-box builders across the East Coast.[7] That background gave Donahue an unusually granular understanding of where the process broke down for buyers — the opacity of builder selection, the fragmentation of design and construction, the absence of any single coordinating layer that a homebuyer could trust. He dropped out of NC State and moved to the Bay Area before founding the company.[7]
Donahue co-founded Atmos with Trent Hedge, Matthew Rastovac, and Austin Kahn, with Crunchbase placing the founding year at 2018.[1] The team's initial concept was VR-focused — an immersive visualization tool for homebuilding — before pivoting to the managed marketplace model that would define the company's public identity.[8] The details of the original VR product and the precise reasons for abandoning it are not documented in any public source, but the pivot suggests the team recognized early that visualization alone was not a defensible business — the value had to sit in the transaction itself.
By the time Atmos entered Y Combinator's Summer 2020 batch — the first fully remote cohort, shaped by the COVID-19 pandemic — the core thesis had crystallized: homebuilding was a $400 billion industry still coordinated largely by phone calls, paper contracts, and fragmented local relationships.[9] Atmos would sit above that chaos as a managed marketplace, providing buyers with a single interface and builders with a qualified, pre-vetted client pipeline.
The founding team's domain credibility was a genuine asset in early fundraising. Sam Altman and Adam Nash participated in the March 2020 seed round, and Khosla Ventures followed within months.[10] The investor base was consistently design-tech and AI-forward — Figma CEO Dylan Field joined the Series A cap table — suggesting the original vision was always software-first, even as execution forced a services-heavy detour.
Rastovac later became Head of Engineering at Atmos before departing to become CEO of Respell, which was subsequently acquired by Salesforce.[11] Hedge went on to found Pylon.[12] The exact timing of these departures relative to Atmos's wind-down is not publicly documented.
Atmos's core product was a digital interface for a process that had never been digitized end-to-end: designing and building a custom home from scratch.
A prospective buyer would enter the platform, specify a location and rough floor plan preferences, and Atmos would match them with vetted local builders from its marketplace network. From there, the platform guided users through a structured design process — selecting layouts, finishes, and exterior styles — using 3D visualization tools to allow clients to approve designs before construction began.[18] Atmos also coordinated construction financing, positioning itself as the single point of contact across what is normally a fragmented set of relationships: architect, builder, lender, and project manager.
The critical architectural detail — and the one that would prove fatal — was that Atmos kept human designers on staff to work directly with clients throughout the design process.[19] Software handled back-end coordination tasks, but the client-facing design work required trained people. This was not a bug the team was unaware of; it was a deliberate choice to ensure quality in a high-stakes, high-dollar transaction. The problem was that it made every engagement labor-intensive and difficult to scale.
Atmos targeted homes in the $400,000–$800,000 range — a price point that sits above entry-level production housing but below the fully custom luxury segment.[14] The focus markets — Raleigh-Durham, Charlotte, Atlanta, Denver, and Austin — were chosen for their combination of high in-migration, available land for new construction, and active local builder ecosystems. These were not markets where buyers were purchasing teardowns in dense urban cores; they were suburban and exurban buyers who wanted something more personalized than a production tract home but lacked the resources or patience for a traditional custom build.
Donahue positioned Atmos explicitly against competitors like Homebound by emphasizing the pre-construction design experience: "We're more design-oriented, and focus more on the process that someone goes through to create the house."[20] Where Homebound emphasized tech-enabled construction management, Atmos emphasized the design journey — a distinction that made sense as a product differentiator but also meant Atmos was investing heavily in the longest, most uncertain phase of the homebuilding funnel.
The product evolved modestly over time. The 3D design tooling became more sophisticated, and the platform expanded its builder network across target markets. But the fundamental architecture — human designers supported by software — did not change materially between the 2020 launch and the 2025 shutdown. No public evidence suggests Atmos attempted a pivot to a pure SaaS model or a builder-facing tool before closing.
Atmos's primary customer was a specific demographic: a first-time custom homebuilder, typically a millennial household with dual incomes, relocating to or within a Sun Belt metro, who wanted more design control than a production builder offered but lacked the time, expertise, or network to manage a traditional custom build. The $400K–$800K price range implied buyers with household incomes likely in the $120K–$200K range — a segment large enough to be meaningful but financially sensitive to interest rate movements.[14]
The secondary customer was the builder: small to mid-size regional contractors who lacked marketing infrastructure and relied on referrals. Atmos offered them a qualified, pre-vetted client pipeline in exchange for a $20,000 flat fee per completed match.[21]
The U.S. residential construction market is large by any measure, with new single-family home starts historically running at 800,000–1,000,000 units per year. The custom and semi-custom segment — Atmos's addressable slice — represents roughly 20–25% of that volume, or approximately 200,000 homes annually. At an average transaction value of $600,000 and a blended take rate of roughly 8–9% (5% buyer fee plus $20K builder fee), the theoretical revenue per completed home was $48,000–$74,000. The addressable revenue pool was large; the question was always throughput.
Atmos competed in a category that attracted significant capital in the 2020–2022 period, when low interest rates made custom homebuilding aspirationally accessible to a broad middle-market demographic.
The most direct competitor was Homebound, which raised $70 million in 2022 and described itself as a "tech-enabled homebuilder."[22] Homebound's model was more vertically integrated — it employed construction managers directly rather than acting as a marketplace — which gave it more operational control but also higher fixed costs. Atmos differentiated on design orientation and pre-construction experience, ceding the construction management layer to its builder partners.
The more structurally important competitive dynamic was not between Atmos and Homebound, but between Atmos and the incumbent production builders — D.R. Horton, Lennar, PulteGroup — who were simultaneously investing in digital design tools and semi-custom product lines. These incumbents had three natural advantages Atmos could not replicate: existing builder relationships at scale, proprietary land pipelines, and the ability to absorb design costs across thousands of units. When Lennar board member Scott Stowell joined Atmos's cap table in 2020, it was a signal of strategic interest — but it also illustrated the risk that incumbents could absorb the design-tech innovation Atmos was pioneering.
The competitive axis that mattered most was not product depth versus distribution reach — it was resilience to rate sensitivity. Atmos's model required buyers to commit to a months-long design process before construction financing was locked. Incumbent production builders, by contrast, offered buyers a completed or near-completed product with a known price and a rate buydown program. When mortgage rates rose from 3% to 7%+ between early 2022 and late 2023, production builders could adjust pricing and offer incentives in real time. Atmos's pipeline of in-progress design clients had no equivalent escape valve.
Atmos operated a double-sided marketplace with fees on both sides of the transaction. Homebuyers paid a 5% service fee on the total cost of construction, covering due diligence, design, and project management. Builders paid a flat $20,000 fee per completed client match.[21]
On a $600,000 home — the midpoint of the target range — the blended revenue per transaction was approximately $50,000 ($30,000 buyer fee plus $20,000 builder fee). This is a high per-transaction figure, but the model's economics depended entirely on throughput. With only 6 completed homes as of November 2022 and approximately 50 over the company's lifetime, the implied lifetime revenue from completed transactions was in the range of $2–2.5 million — a fraction of the approximately $20 million raised.[23]
The $693,750 in revenue booked by February 2021 — roughly nine months post-launch — is consistent with approximately 14–15 completed transactions at the stated fee structure, or a smaller number of higher-value homes.[13] Atmos never disclosed revenue figures after this early data point; the absence of subsequent revenue disclosures, even at the Series A announcement, is itself a signal that growth did not accelerate meaningfully.
The company never disclosed burn rate or headcount. Given the operational model — human designers on staff, project coordination across five markets, a builder vetting function — a reasonable inference is that annual operating costs were well in excess of $3–4 million, implying the $12.5M Series A provided roughly 3 years of runway at best. The timeline from Series A (November 2022) to shutdown (~March 2025) — approximately 28 months — is consistent with that estimate.
Atmos's traction data tells a story of a wide design funnel with a very narrow construction conversion rate.
At the October 2020 fundraise — five months after launch — the company reported "the first dozen homes" being designed and one home having broken ground.[15] By November 2022 — two full years later and after $12.5M in fresh capital — Atmos had completed only six homes and was "working on a few dozen more."[4] That is a completion rate of roughly three homes per year across five target markets over the company's most active growth period.
Over its full lifetime, Atmos designed over $200 million in homes and completed approximately 50 builds.[5] The gap between $200M designed and 50 homes built — at an average of $600K per home, 50 completions represents $30M in construction value — implies a design-to-build conversion rate of roughly 15%. The remaining 85% of design work generated no construction revenue, only cost.
External recognition was strong: Forbes 30 Under 30 for Donahue and Hedge in 2021,[16] and the Charlotte Inno Startups to Watch award.[24] These validated the team's credibility but masked the throughput problem that would ultimately prove fatal.
The Federal Reserve raised the federal funds rate from 0.25% in March 2022 to 5.25–5.50% by July 2023 — the fastest tightening cycle in four decades. Thirty-year fixed mortgage rates, which had averaged 3.1% in 2021, exceeded 7% by late 2022 and remained elevated through 2023. For a buyer financing a $600,000 home, this represented a monthly payment increase of approximately $1,500–$2,000 — enough to push the transaction out of reach for a large share of Atmos's target demographic.
The damage to Atmos was not just that new customers stopped entering the funnel. The more acute problem was that customers who had already spent months in the design process — and in whom Atmos had already invested significant designer time and coordination effort — could no longer afford to proceed. Donahue confirmed this directly: the rate increases caused clients "who had spent months designing their dream homes to no longer be able to afford them."[17] The design work was sunk cost; the construction revenue never materialized.
Atmos attempted to address this by raising the $12.5M Series A in November 2022 — precisely when the rate environment had already deteriorated significantly. The raise extended runway but did not change the underlying dynamic: the product required buyers to commit to a multi-month, high-cost process in an environment where the end-state financing was increasingly uncertain. No evidence suggests Atmos introduced rate buydown programs, adjusted its target price range downward, or pivoted to markets with lower construction costs.
Donahue's own retrospective is unusually candid: Atmos became "this extremely operational business" and "kind of like a glamorized architecture firm," and "we never quite replaced the humans."[19] He further conceded that the business "wasn't scalable or resilient."[6]
The original thesis was that software would progressively automate the design and coordination workflow, reducing the human labor required per transaction and improving margins as volume scaled. This is the standard managed marketplace playbook: use humans to establish quality standards, then replace humans with software as patterns emerge. Atmos never completed that transition. The 3D design tools handled visualization, but the judgment-intensive work of translating a client's preferences into a buildable, code-compliant design remained human.
The consequence was a cost structure that scaled linearly with volume rather than sublinearly. Each new client required roughly the same designer hours as the last. At 50 homes over five years — roughly 10 per year — the operation could not generate enough revenue to cover its fixed costs, let alone invest in the automation that might have changed the unit economics.
The team attempted to address this by raising capital to fund continued operations while presumably investing in product development. But the gap between the company's stated vision (a one-click homebuilding platform) and its operational reality (a staffed design service) was never closed. The Series A announcement in November 2022 described the platform in terms consistent with the 2020 launch — no evidence of a material product evolution that reduced human dependency.
Most businesses are affected by interest rate changes through their cost of capital. Atmos was affected through its customers' cost of capital — a more direct and more dangerous exposure. Custom homebuilding requires construction financing (a short-term, rate-sensitive product) and permanent mortgage financing (a long-term, rate-sensitive product). Both became dramatically more expensive in 2022–2023.
More specifically, Atmos's $400K–$800K target range sits in the segment where rate sensitivity is most acute. Buyers in this range are typically stretching to afford the purchase; they have less financial cushion than luxury buyers and less flexibility to absorb payment increases. Production builders in this segment responded to rate increases by offering mortgage rate buydowns and price reductions — tools that required scale and balance sheet capacity Atmos did not have.
The structural lesson is that a marketplace whose transaction value is determined by a rate-sensitive financing product is inherently cyclical. Atmos's business model was viable in a low-rate environment and structurally impaired in a high-rate one. The company raised its largest round at the worst possible moment in the rate cycle, suggesting either that the team underestimated the macro headwinds or that the raise was necessary for survival regardless of timing.
Matthew Rastovac departed Atmos to become CEO of Respell, which was subsequently acquired by Salesforce.[11] Trent Hedge departed to found Pylon.[12] The exact timing of these departures is not documented, but both founders built and led separate companies — suggesting they left with sufficient time and energy to start new ventures, implying departure well before the March 2025 shutdown.
Co-founder departures in a services-heavy business are particularly damaging because institutional knowledge about builder relationships, client management processes, and product architecture is concentrated in a small team. Whether the departures caused the operational difficulties or resulted from them is unknowable from public data, but the signal is the same: by the time Atmos shut down, the founding team had dispersed.
A managed marketplace that never automates its human layer is a staffing agency with a better pitch deck. Atmos raised $20M on the premise that software would progressively replace its human designers and coordinators. By the time it shut down, five years after launch, Donahue's own description — "glamorized architecture firm" — confirmed the automation thesis had not been validated. The lesson is not generic ("automate early") but specific: Atmos's design workflow involved judgment calls — translating client preferences into buildable, code-compliant plans — that 2020–2022 technology could not reliably automate. Donahue's successor company, Drafted, launched in December 2025 with AI-generated floor plans and no designers on staff, suggesting the enabling technology arrived only after Atmos had already consumed its capital.
Raising a large round at the peak of a rate cycle does not insulate a rate-sensitive business — it extends the runway into the headwind. Atmos's $12.5M Series A closed in November 2022, when the Fed had already raised rates by 375 basis points and mortgage rates had already crossed 7%. The capital extended operations for approximately 28 months but did not change the fundamental problem: the target customer could no longer afford the product. A smaller, faster pivot — to lower price points, to builder-facing SaaS, to markets with lower construction costs — might have been more valuable than a large raise that funded the existing model through a hostile environment.
A design funnel that converts at 15% is a marketing cost, not a business. Atmos designed over $200M in homes and completed approximately 50 builds — implying roughly $30M in construction value actually reached completion against $200M in design work initiated. Every design engagement that did not convert to construction represented sunk designer time, coordination cost, and opportunity cost. Atmos never publicly disclosed its funnel conversion rate, which is itself a signal. A business with healthy conversion metrics discloses them; the absence of this data from every funding announcement suggests the team knew the ratio was unfavorable.
Targeting a demographic that is maximally rate-sensitive at a moment of historically low rates is a timing bet, not a market insight. Atmos's $400K–$800K buyer in Raleigh or Charlotte in 2020 was a plausible customer when 30-year rates were 3%. The same buyer in 2023, facing 7%+ rates, faced monthly payments $1,500–$2,000 higher on the same home. Atmos's market selection — Sun Belt metros, middle-market price points, first-time custom builders — was well-reasoned for the 2020 environment and structurally fragile in the 2022–2023 environment. The company had no product or pricing lever to adjust for this shift.
Strategic investors signal credibility but not resilience. Atmos's cap table included Khosla Ventures, Sam Altman, JLL Spark, a Lennar board member, Keller Williams, and Figma's CEO — a roster that validated the company's design-tech positioning and real estate industry relationships. None of these relationships translated into distribution advantages, builder network effects, or financing products that could have cushioned the rate shock. Strategic investment is valuable for market access; it does not substitute for a business model that can survive a macro cycle.