Building a Climate Tech Startup
Discussion w/ Panos Stravopodis - Elyos Energy co-founder, exit to OpenAI, entrepreneurial resilience, AI product managers, jobs, and more
This is Startup Pirate #101, a writing about technology, entrepreneurship, and startups every two weeks. Made in Greece. If you haven’t subscribed, join 6,200 readers by clicking below:
Building a Climate Tech Startup
The following is a conversation with Panos Stravopodis, co-founder and CTO of Elyos Energy, a climate tech startup that decarbonises commercial and industrial buildings using AI. They do so by optimising energy consumption and playing a significant role in the operation of the power grid — that intricate system of lines, cables and transformers which provide electricity all the way from its generation to customers. We covered a number of topics:
Why the power grid is broken
Energy flexibility and the importance of balancing energy demand and supply
Unlocking energy flexibility for commercial and industrial buildings
Launching early for fast validation of ideas
Investing in frequent and reliable software delivery
Building things that don't scale
and much more.
Pano, it's great to finally talk to you on the record! There are a few things I want to cover today, so I'll jump right in. Your co-founders and you had backgrounds in energy. Why did you decide to focus on commercial and industrial buildings' costs and emissions out of the entire spectrum of energy problems?
PS: Alex, I've been a fan of this newsletter, so I'm glad to be here and share our journey with Elyos. Pip, Adrian, and I knew each other and had been experimenting with new startup ideas for a while. We all had prior experiences in energy. Pip was head of strategy at OVO Energy, one of the biggest green energy suppliers in the UK, and Adrian led several projects for large customers at the Energy Practice of Boston Consulting Group. I was looking after engineering at Bulb Energy, a fast-growing renewable energy company.
While still in the ideation phase, we looked at a better financing model for solar farms. That's when we realised the biggest barrier to more renewables is not access to capital. Last year, more than £200bn worth of projects were sitting in the connections queue in the UK. The biggest barrier is that it currently takes over 10 years to connect to the grid. The grid is broken.
We believe the answer is energy flexibility, which is the ability to distribute the existing energy capacity in a smarter way instead of constantly generating more supply than demand to ensure security of electricity in peak periods. The latter is just not feasible, given the electrification of everything, the intermittency of renewables, and the insufficient grid infrastructure stalling the pace of net zero.
Plenty of projects address flexibility for residential customers, although that segment represents only 35-40% of the energy usage. I worked on one of these while still at Bulb, e.g. using time-of-use tariffs for electric cars/chargers and battery optimisation for home installations. While iterating on different ideas at the beginning of 2023, we realised that the commercial and industrial space, which takes up the other 60%-65% of energy usage globally, was underserviced due to the complexities of customers (different vendors per country, complex systems, time-sensitive equipment, security concerns, different requirements per site, etc). We decided this had to change.
You highlighted energy flexibility, so I want to double-click here to better understand its importance and scope.
PS: Energy flexibility is about balancing demand and supply in the energy system in response to an external signal (say, a change in price which might be correlated with renewables generation at a given time of the day). Essentially, it’s the ability to shift energy usage to lower-priced hours or respond to events from the grid to increase or decrease consumption for a period of time. This becomes vital in the context of the following waves:
Electricity consumption is expected to triple in the next 20 years as a response to the electrification of heat, rapid adoption of electric vehicles, and growing demand for air conditioning.
The rise of intermittent sources of energy such as wind, solar, etc., which are green but their supply is unstable, e.g. not enough wind, cloudy days in countries like the UK, or even too much sun in Greece during summer overheating cables and leading to electricity shutdowns.
An important enabler of it is the rise of distributed storage such as batteries that are relatively cheap and can store amounts of energy which will be later used by users or discharged to the grid. A common example of energy flexibility in residential sites is consumers using apps to monitor electricity prices and scheduling their dishwashers to run midmorning rather than when demand and prices are at their highest.
Grid operators need to adjust their setup daily based on the demand and optimise for the set frequency to offset any imbalances from the mismatch of demand and supply. Currently, we either use interconnectors or gas-powered factories to compensate for such imbalances. In fact, the UK grid spends $4bn annually leveraging gas peakers in an attempt to reduce the risk of blackouts during periods of peak electricity usage.
So, if we want to decarbonise, the only way is energy flexibility at scale. And to achieve this, the path passes through commercial and industrial buildings.
So, how do you unlock energy flexibility for these sectors of the economy?
PS: Our technology has three pillars: energy visibility, optimisation, and flexibility. We focus on commercial and industrial buildings, typically over 1,000 sq m., from office spaces to universities and local authorities to malls and hotels.
We start by connecting to the buildings' smart meters and submetering solutions, building management systems, heating, ventilation, and air conditioning (HVAC), solar inverters, and batteries. We do that through APIs or, when this is not possible, by installing Elyos Gateway, our proprietary technology that allows to connect to all building automation systems.
This is an IoT Gateway that supports all major BMS protocols and enables us to integrate seamlessly into submeters, solar inverters, storage systems, etc., that might not support smart functionality or whose APIs are locked by vendors. All in all, the first step is to help our customers understand how their buildings operate and visualise energy consumption across their sites.
After connectivity is established, we start optimising energy consumption. That comes in two modes and two verticals:
Read-only, where our models provide insights and recommendations to customers while continuously doing anomaly detection in the background.
Read-write, where we take complete control of the building and optimise the energy usage for our customers.
Building level, taking into consideration features like weather, building occupancy, pricing, and battery utilisation.
Asset level, e.g. per HVAC system or battery optimising their performance.
The third pillar is energy flexibility. After we have control of the building, we can easily shift energy usage outside peak demand periods, helping save cost (using flexible energy tariffs, aka Time-of-Use tariffs) and carbon. For example, we can charge up a building's battery systems when energy demand is low and power its HVAC through that stored energy instead of consuming from the grid when demand is at its highest. Or, we can turn on the air conditioning at different times during the week, adjusting ventilation accordingly and using heat produced by the chillers for heat exchange systems. We can apply similar techniques to boilers, and a big opportunity we see with some customers is electric chargers.
At the same time, we enable customers to participate in local flexibility markets, where they can trade flexibility in energy usage to balance supply and demand in real time, contributing to grid stability and efficiency. This can occur implicitly (proofs from metering systems that they reduced consumption when energy demand was high) or explicitly (committed, dispatchable resource that can be traded on various energy markets).
That creates an additional revenue stream for real estate managers as grid operators pay users who offer flexibility when needed. We started seeing this trend 20 years ago with the deregulation of energy markets in Europe, and today, it's widespread, with most countries having multiple local markets that were streamlined last year by EU legislation and are currently being migrated to use these standards.
I'd love to dive deep into your MVP. How did you strike a balance between launching early and iterating vs. perfecting your technology before getting it into users' hands?
PS: One of the first things they told us at YC was to launch as early as possible, allowing for fast validation, and that's what we did. We started reaching out to potential customers from the early days and built "design partner" relationships with real estate owners, building and facility managers, finance managers who needed support with planning and forecasting, operators with one or multiple assets, etc. We hacked day and night, and given we were building the MVP while in San Francisco and most of our customers were in Europe, we would ship new functionality overnight and validate ideas in a few hours.
What might sound non-consensus is that we invested very early on in CI/CD pipelines for frequent and reliable software delivery. This helped us push thousands of lines of code daily. We were integrating smart meters and building facilities, thereby "commanding" devices used by real people in offices, malls, hotels, etc., so we wanted to ensure the code we shipped was thoroughly tested.
Like every startup, we also built things that don't scale. For instance, we have a feature that scrapes data from energy bills using OCR (technology that recognises text within a digital image) and LLMs. When we first launched it, the functionality was basic and not particularly accurate, which meant that we had to manually enter data from thousands of customer bills. The reality is customers don't care about the tech running behind the scenes; they want their problems solved.
What are the next steps for Elyos Energy?
PS: We currently work with customers in the UK, Netherlands, Poland, Denmark, Greece, and the US and operate in hundreds of buildings. Our primary focus is to continue expanding internationally and ensure we have our customers fully up and running in no time. Hence, a key part of that will be remote deployments, as, in some cases, customers might be in hard-to-reach places.
We are also doubling down on fine-tuning our machine learning models to cover different building profiles and countries while working on the next version of our Elyos Gateway. Enabling commercial and industrial customers with flexible loads will have a massive impact on the efficiency of power grids and the decarbonisation of the economy, and we couldn't be more excited about the times ahead.
Thank you, Pano; I appreciate it!
PS: Thanks, Alex.
Jobs
Check out job openings here from startups hiring in Greece.
News
AI platform to accelerate computer vision in edge devices Axelera announced a $68m Series B. I wrote more about their technology in a past issue — jump here to learn more.
Collaboration tool for MacOS Multi was acquired by OpenAI.
Blockchain data startup Openstory was acquired by Ava Protocol, an EigenLayer actively validated service.
hoper launched the first helicopter airline in Greece with €3m in funding.
Seafair secured an additional $1.5m led by FJ Labs to help shipping companies manage data and recruit seafarers.
edTech platform Workearly raised €1.1m led by Metavallon VC.
Aristotle University of Thessaloniki spin-off Athletopia secured €180k to supercharge running events.
Trade-in platform for smart devices Pandas secured funding from SPOROS.
Resources
National Technical University of Athens and University of Patras putting Greece on Europe's top-tier AI talent map (from MacroPolo's global AI talent tracker).
Building entrepreneurial resilience with Antonis Kalipetis and Paris Kasidiaris at Mikri Kouventa.
Latest report by Sifted on the growing Greek tech scene.
The shipping fees saga by Vasso Kalaitzidou, Head of Category Management at Skroutz.
Navigating the shipping industry with Antonis Malaxianakis, founder & CEO at Harbor Lab.
AI Product Management: levels, tasks & scope by Marily Nika, Gen AI Product Lead at Google.
Fallacies engineers make in architecture decisions by Panayiotis Kritiotis, Senior Engineering Manager at Linearity.
On AI ethics and regulation by Konstantine Arkoudas.
Events
“26th Open Coffee Heraklion” on Jun 28
“The foundations of modern AI” by Archimedes Research Unit on Jul 3-4
“Athens NLP 2024 Summer School” on Sep 19-25
That’s all for this week. Tap the heart ❤️ below if you liked this piece — it helps me understand which themes you like best and what I should do more.
Find me on LinkedIn or Twitter.
Thank you for reading,
Alex