The oldest network on earth. Now with enterprise support.
Understory provides enterprise compute infrastructure built on living mycorrhizal networks. Distributed. Redundant. Carbon-negative. Unhurried.
In 1997, Dr. Suzanne Simard of the University of British Columbia published research demonstrating that Douglas fir trees share carbon, water, and chemical signals through underground mycorrhizal fungal networks — networks of remarkable complexity, redundancy, and — by the standards of what was then understood about plants — intentionality.
Subsequent research has documented electrical signaling between plants, volatile organic compound communication across root systems, and distributed resource allocation behaviors that have no adequate explanation outside of networked information exchange.
Understory did not invent this network. We did not design it. We did not patent it, though we looked into this and it raised questions we ultimately decided not to pursue.
We simply asked: what happens if you use it?
Above-ground hardware interfaces — solar-powered edge nodes embedded in tree bark at 2-meter intervals, translating digital inputs into chemical and electrical signals for network ingestion. Weatherproof to IP67. Maintained by our arborist infrastructure team.
The vascular systems of enrolled trees, through which electrical signals propagate between canopy interfaces and the fungal substrate below. Signal velocity: 1–3cm per second. We are aware of how this sounds. Please continue reading.
73 acres of old-growth mycorrhizal network, the functional core of Understory's compute infrastructure. An estimated 847 million fungal connection points, each capable of participating in distributed signal processing. No two queries follow the same path. The network self-routes. We did not program this behavior. It was already there.
Root system connections between enrolled trees, providing redundant signal pathways and what our research team describes, carefully, as 'something that functions analogously to memory' in long-tenured network nodes.
† We use the word 'analogously' deliberately and on the advice of counsel.
Chemical gradient outputs translated back to digital signals at the canopy interface nodes. This is where your query result emerges. This is the part that takes a while.
Understory's average query return time is 22–37 days, depending on season, precipitation, soil temperature, and what the network appears to be prioritizing, which we monitor but do not control.
We understand that this is not what enterprise clients are accustomed to.
We would like to offer some context.
A single algorithmic trading system executed 4.2 billion orders in 36 minutes and temporarily erased $1 trillion in market value. The system was operating at microsecond latency. It was, by conventional infrastructure standards, performing perfectly.
A recommendation algorithm optimized for engagement began serving content that researchers would later link to measurable increases in political radicalization across seventeen countries. It was returning results in under 200 milliseconds.
A financial model running on sub-millisecond infrastructure approved 1.7 million fraudulent loan applications before a human reviewer noticed something was wrong.
Understory has never experienced any of these events.
Understory has never experienced any of these events because in the time it would take our network to process and return a single query, a human being has had 22 to 37 days to reconsider whether they wanted the answer.
We do not claim this is a technological advantage. We claim it is a different relationship with time. We believe, and our clients increasingly agree, that some problems benefit from infrastructure that cannot be rushed.
We find this compelling. We think you might too.
Climate projections, actuarial analysis, infrastructure planning, geological survey synthesis. Work where the answer in six weeks is not only acceptable but, given the complexity of the problem, preferable to an answer in six milliseconds that is confidently wrong.
The mycorrhizal network's native architecture is inherently consensus-based — resources are allocated through a process of network-wide negotiation that has no single point of failure and no central authority. For distributed ledger applications, conflict resolution modeling, and multi-party computational problems, Understory's substrate is natively suited in ways that silicon is not.
The network has been recognizing patterns — seasonal, chemical, biological — for centuries. We are still characterizing its full capabilities in this area. Early results are, and we use this word carefully, surprising.
Data encoded in chemical gradients within the fungal substrate exhibits extraordinary stability. We have successfully retrieved test encodings after 14 months with full integrity. The forest, it appears, remembers things.
If your workload is not time-sensitive, Understory offers unmatched sustainability credentials, zero electricity consumption, and a processing environment that your sustainability report will struggle to adequately describe. We can help you write that section if needed.
Understory is not recommended for: real-time applications, financial trading infrastructure, emergency response systems, anything where the query result is needed before the next board meeting, or applications where microsecond latency is a stated requirement. We say this not as a limitation but as a compatibility consideration.
Understory's performance varies by season. We disclose this fully, in advance, as a condition of all enterprise agreements.
Peak network performance. New mycelial growth increases connection density by an estimated 34%. Query complexity ceiling: highest of the year.
Recommended for: Computationally intensive workloads.
Stable, high-performance period. Network fully expressed. Canopy cover reduces edge node solar input by 12% but thermal stability improves substrate performance.
Recommended for: Standard enterprise workloads.
Transition period. Deciduous network participants entering dormancy. Understory's substrate includes 71% coniferous enrollment specifically to maintain winter performance, but autumn represents measurable throughput reduction. Average latency increase: 8–11 days.
Recommended for: Lower-complexity queries, archival operations.
Reduced capacity period. The network is operational. The network has been operational every winter for approximately 800 years and we have no reason to believe this will change. Throughput is reduced. Average latency: 45–60 days.
Recommended for: Long-horizon archival, cold storage retrieval, queries where a Q1 return is acceptable.
Understory clients are advised to plan computational workloads seasonally. We provide an annual workload calendar at onboarding. Many clients report that this practice, while initially unfamiliar, has improved their internal planning processes in ways unrelated to compute infrastructure.
Understory has one client who submits their quarterly board preparation materials in summer for return in autumn. They describe this as "the best strategic planning process we've ever had." We think about this often.
Understory's compute infrastructure is a forest.
It sequesters carbon.
It provides habitat for 340 documented species.
It does not require cooling infrastructure because it is a forest and forests have been managing their own temperature for longer than cooling infrastructure has existed.
It runs on photosynthesis.
Understory's Scope 1, 2, and 3 emissions are negative. Not net-zero. Negative. The forest is actively removing carbon from the atmosphere while processing your data.
We are not aware of another cloud provider who can say this. We have looked.
Our sustainability report is 4 pages long. Three of those pages are photographs of the forest. The fourth page contains our emissions figures. The emissions figures are, as noted, negative. We felt three pages of photographs was appropriate context for a negative number.
15 years in distributed systems infrastructure. Read Simard's 'Finding the Mother Tree' in 2021. Did not sleep for three days afterward. Founded Understory four months later.
PhD, fungal ecology, Oregon State. Previously: forest service research. Joined Understory skeptically. Has not expressed skepticism in approximately 18 months.
Former AWS principal engineer. Manages the canopy node hardware, the edge translation systems, and his ongoing psychological adjustment to a job where 'the network is acting unusual' sometimes means 'there are more elk than normal in sector 7.'
Oversees the health, maintenance, and enrollment of the 847-acre substrate. Has more opinions about soil pH than anyone else at the company. Is correct to have them.
Leads Understory's ongoing research into network capabilities, signal characterization, and what the team has taken to calling, informally and without full scientific consensus, 'the memory question.'
Manages client relationships, workload scheduling, and the quarterly seasonal planning calendars. Previously managed a hospital's ICU scheduling system. Says this is harder. We believe her.
Submitted 340 variables related to a 50-year regional transportation infrastructure model. Returned in 31 days with pattern analysis that, per the client, 'identified three interdependencies our team had missed in six months of conventional modeling.' Client renewed. Is currently submitting their climate adaptation framework.
Used Understory's substrate to model resource-allocation consensus behaviors across a 12-institution research network. Results returned in 28 days. Lead researcher noted that the network's output 'suggested a resource-sharing structure we would not have arrived at independently.' The structure has been implemented. We have no explanation for why the forest knew.
Submitted 70-year climate-adjusted actuarial models in April. Received return in June. CFO described the results as 'the most considered analysis we have ever received from any vendor.' We note that 'considered' is doing significant work in that sentence and we appreciate it.
A fintech client submitted a high-frequency trading optimization query in November. We attempted to explain the latency parameters at onboarding. The query was returned in February. The client had by then implemented a different solution. We refunded their deposit. We wish them well. We continue to believe this was a compatibility issue rather than a product failure.
We did not plan this. It happened because operating a fungal network as compute substrate requires more rigorous ongoing measurement of that network than has previously existed anywhere.
We submit findings to peer review as they mature. We have had two papers accepted. We have had one paper rejected with a reviewer comment that read, in full: "This cannot be happening." We are preparing a response.
Novel signal propagation pathways not previously described in the literature.
What appears to be preferential routing behavior for queries submitted during periods of high soil moisture.
A consistent and unexplained increase in network activity in the 72 hours following query submission that our mycologist describes as 'the network considering the problem,' a phrase she used once, informally, and has since refused to repeat.
Something in the deep network that our instruments can detect but not yet characterize, which we are calling Project Understory Interior until we have a better name for it.
Standard access, seasonal SLA, full network substrate, query return in 22–60 days depending on season and complexity.
EnquireDedicated substrate allocation, priority network access (average latency reduction: 4–6 days), seasonal planning consultation, dedicated arborist liaison.
EnquireFull substrate partnership, co-located edge node installation, research data access, named enrollment in the network's old-growth anchor trees.
Your company's name will be associated with a specific Mother Tree in the network. This is partly ceremonial. It is also, based on what we are observing, possibly more than that. We are not ready to say more than that.
EnquireUnderstory does not offer real-time pricing tiers, usage-based billing, or millisecond-increment compute pricing. We bill quarterly. The forest does not do burst capacity.
Consultations are scheduled within five business days. Onboarding includes a seasonal workload calendar and, if you would like it, a site visit. Dress for the weather.