At Leveiir Capital, the beginning of each year prompts a deliberate exercise in synthesis: we review every investment decision we made in the prior year, every company we passed on and why, every market hypothesis we tested and what the evidence showed, and we try to produce a clear-eyed account of what we are seeing that is useful not just for our own investment process but for the founders and fellow investors in our network. This is that account for 2025.
The five themes we discuss below are not predictions about specific companies or technologies. They are observations about structural forces that we believe will shape where the most important seed-stage opportunities sit over the next two to four years. Some of these themes will be familiar; our goal is to add texture and specificity to the familiar, and to identify the specific implications that are less widely understood.
Theme 1: The AI Infrastructure Buildout — The Real AI Money Is in the Stack Below the Models
The dominant narrative in AI investment since 2023 has focused on foundation models — GPT-4, Claude, Gemini, and their successors. This narrative is not wrong; foundation models represent a genuine technological breakthrough that is reshaping the economics of software. But the investment opportunity in foundation models themselves has largely been captured by a small number of very large, very well-funded organizations. For seed-stage investors, the more interesting opportunity lies in the layer below the models: the infrastructure required to make AI useful, reliable, and economically viable in enterprise production environments.
We are thinking about AI infrastructure across four sub-categories. First, data infrastructure: the pipelines, labeling tools, quality management systems, and synthetic data platforms that enterprises need to train and fine-tune models on proprietary data at scale. Most enterprises want to use AI on their own data — their financial records, their customer interactions, their operational data — and the tooling required to do this safely and effectively is dramatically underdeveloped relative to the demand. Second, evaluation and testing: as AI systems are deployed in high-stakes applications — clinical decision support, financial risk assessment, legal research — the ability to measure and certify their accuracy and reliability becomes critical. The market for AI evaluation tooling is in its earliest innings, and the companies building robust evaluation frameworks for specific high-stakes domains will have enormous value in the next three to five years.
Third, deployment and MLOps: the gap between a working AI model in a research environment and a reliable AI model in a production enterprise environment is enormous, and most enterprises lack the in-house expertise to bridge it. The platforms that abstract away this complexity — handling versioning, monitoring, rollback, and integration with existing enterprise systems — are building deeply sticky, high-value infrastructure. Fourth, AI security: as AI systems take on more autonomous roles in enterprise workflows, the attack surface for AI-specific vulnerabilities (prompt injection, model poisoning, output manipulation) expands in ways that conventional cybersecurity tools were not designed to address. AI security is one of the most underfunded subcategories we follow, and we believe it is one year away from becoming a mainstream enterprise priority.
For founders: the companies we want to meet in this space are building tools that solve specific, verifiable problems in enterprise AI deployment — not tools that make AI generally "better," but tools that solve the specific friction points that are preventing AI from moving from pilot to production in regulated industries, large enterprises, and government environments.
Theme 2: Sovereign Technology — Every Government Wants to Control Its Critical Infrastructure
One of the most significant structural shifts in the global technology landscape over the last three years is the emergence of what we call sovereign technology demand: the growing insistence of governments and large state-adjacent institutions that critical technology infrastructure be operated from within their jurisdictions, with data residency in their territories, and with the ability to audit and control the underlying code and algorithms. This trend is visible across cloud infrastructure (EU data sovereignty regulations, DORA for financial institutions), AI governance (EU AI Act, national AI safety bodies), and communications (the ongoing fragmentation of global internet infrastructure along geopolitical lines).
For technology founders, this creates a specific and highly monetizable opportunity: building the compliance, security, and infrastructure tooling that enables existing technology products to operate within sovereign constraints. This is not a glamorous category, but it is an extraordinarily large and growing one. Every enterprise software vendor operating in the EU, UK, or any of the growing number of jurisdictions with meaningful data sovereignty requirements needs to either build or buy this capability — and most will buy it.
The companies we are most interested in within this theme are those building sovereign-ready cloud infrastructure (private cloud and on-premise deployment tools that make cloud-native applications deployable in sovereign environments without re-architecture), regulatory compliance automation (AI-powered tools that continuously monitor regulatory requirements and map them to operational policies — a problem that grows in complexity faster than any team can manage manually), and cryptographic infrastructure (the encryption, key management, and secure computation tools that underpin sovereign data handling). ComplianceAI, one of our portfolio companies, is building in this last space for the EU financial services sector, and their growth trajectory reflects the urgency of the demand.
Theme 3: Climate as a Technology Platform — The Energy Transition Is a Software Problem
The energy transition is the most frequently discussed investment theme of the decade, and for good reason: the scale of the required infrastructure buildout is genuinely unprecedented. But the narrative around climate tech investment has historically focused heavily on hardware — batteries, electrolysers, solar panels, heat pumps. We believe this focus is correct for climate impact but incomplete for venture-scale return generation, because hardware businesses at the energy transition scale require project financing structures, balance-sheet capital, and development timelines that are not well-suited to the venture capital model.
The more interesting opportunity for seed-stage technology investors is in the software and intelligence layer that makes the energy transition work. The energy transition is generating an enormous amount of new data — from smart meters, from grid sensors, from satellite monitoring of emissions, from energy markets, from supply chains — and the software to make sense of this data, to optimize decisions across it, and to satisfy the reporting requirements that policymakers are imposing on energy operators and large enterprises is dramatically underdeveloped relative to the scale of the data.
We are specifically watching three sub-themes. Carbon intelligence: the regulatory requirements for scope 1, 2, and 3 emissions reporting are expanding rapidly, and the data quality problem — particularly for scope 3, which requires tracking emissions throughout the supply chain — is a genuinely hard data engineering and machine learning problem. The companies building rigorous, auditable, data-native emissions tracking platforms are addressing a legal compliance requirement that is growing in urgency every year. Grid intelligence: the transition from centralized fossil fuel generation to distributed renewable generation is creating an enormous optimization problem for grid operators — the need to balance supply and demand in real time across a network that is fundamentally more complex and more volatile than the legacy grid it is replacing. Software platforms that aggregate, analyze, and optimize distributed energy resources are building in a market that will grow for decades. Climate risk: financial institutions, real estate operators, and infrastructure owners are increasingly required to quantify and report on their exposure to physical climate risk — flooding, extreme heat, drought, wildfire. The data science and modeling infrastructure to do this at the asset level is in its earliest stages, and the regulatory mandate driving demand is intensifying.
Theme 4: Embedded Finance 2.0 — Financial Products as a Software Feature
The first wave of embedded finance — enabling software platforms to add payment processing to their products without becoming payment companies — is largely mature. The second wave is substantially more interesting: enabling software platforms to offer complex financial products — working capital loans, revenue-based financing, business insurance, expense management, treasury management — to their captive user bases, without the regulatory complexity of becoming licensed financial institutions.
The opportunity exists at the intersection of three converging trends. B2B software platforms have accumulated enormous amounts of proprietary data about their customers' financial health — revenue patterns, invoice cycles, cash flow dynamics, payment behaviors — that represent a far better underwriting signal than anything available to traditional banks. Regulatory frameworks in the EU and UK have evolved to enable non-bank institutions to offer financial products through banking-as-a-service relationships with licensed deposit-taking institutions, dramatically reducing the regulatory complexity of embedded financial products. And the cost of building financial products on top of modern API infrastructure has fallen to the point where it is economically viable for a small team to build and launch a working capital product in months rather than years.
The companies we are most interested in are building the middleware layer — the compliance infrastructure, risk models, and product templates that enable any B2B software platform to add financial products in weeks rather than months, with the regulatory and risk management complexity handled by the middleware provider rather than the platform. Bankable Flow, one of our portfolio companies, is building exactly this, and their growth — $120M+ in monthly credit decisions within 18 months of launch — reflects the strength of the demand. For founders: if you are building embedded financial infrastructure and you have a view on the specific segment (vertical software, geographic market, or product category) where the regulatory tailwind and data advantage are greatest, we want to hear from you.
Theme 5: Longevity and Preventive Health — The New Economy of Staying Well
The final theme is the one with the longest time horizon and the largest ultimate market size: the convergence of longevity science, digital health, and preventive medicine into a new economic model for healthcare that is organized around maintaining health rather than treating disease. This convergence has been anticipated for years, and has been slow to arrive — but we believe the preconditions for commercial takeoff are now in place in a way that they were not as recently as 2022.
The scientific foundation is more solid than it has ever been. The last five years have seen a series of genuine breakthroughs in our understanding of the biological mechanisms of aging — senolytic compounds, NAD+ metabolism, mTOR pathway modulation, gut microbiome interventions — that have produced compelling preclinical data and, in several cases, early-stage clinical results that were simply not available at the beginning of the decade. The combination of genetic data at population scale, proteomics, and machine learning is enabling the identification of aging biomarkers and intervention targets with a precision that was not achievable with prior scientific tools.
The commercial infrastructure is maturing. Consumer longevity clinics — analogous to the dental or optometry practice model — are growing rapidly in major cities, creating distribution channels for evidence-based preventive health products that did not exist five years ago. Direct-to-consumer diagnostics platforms have reduced the cost of blood biomarker panels to the point where quarterly monitoring by health-conscious individuals is economically practical. And the insurance and employer benefits ecosystem is beginning to recognize that preventive health interventions have quantifiable ROI in terms of reduced acute care costs, which is beginning to open institutional distribution channels.
For founders: the companies we are most interested in at the intersection of longevity and technology are not consumer wellness brands — the market is too crowded and the differentiation too difficult. We are interested in the clinical software infrastructure that enables longevity medicine to be practiced at scale (the EHR and care coordination tools designed specifically for the preventive care model), the diagnostic platforms that can identify actionable health risks years before clinical manifestation, and the precision therapeutics companies developing interventions with strong biological rationale and rigorous clinical validation. TalentLattice and MediThread in our portfolio reflect early expressions of this thesis in adjacent domains; we are actively looking for companies working on the core longevity technology and clinical infrastructure layer.
Looking Forward
The common thread running through all five of these themes is the shift from software as a layer on top of existing systems to software as the connective tissue of entirely new systems — new energy infrastructure, new healthcare delivery models, new financial services architectures, new government-technology relationships. This is a more demanding and more consequential kind of software building than the first two decades of the SaaS era. It requires founders who combine deep domain expertise with genuine technical ability, who think about the regulatory and policy environment as a strategic variable rather than a constraint, and who are building for a global market from the very beginning.
Those are the founders we most want to back in 2025. If you are building at the intersection of any of these themes and believe you have the right insight, team, and timing, we are eager to learn from you. The best investment meetings we have are the ones where we leave knowing something we did not know before. We are always looking for the next one.