What Is Horizontal AI?
Horizontal AI refers to general purpose artificial intelligence systems designed to perform broad, cross industry functions rather than being specialized for a single sector or task. Built on expansive foundation models, these systems excel at natural language processing, data analysis, content creation, and workflow automation across virtually any use case.
Their advantage is scalability, cost effectiveness, and speed of deployment because they are pre trained on vast, diverse datasets, organizations can adopt them out of the box without building custom infrastructure for every function. The tradeoff is depth. They lack the built in domain expertise or strict regulatory customization that specialized work often demands.
What Is Vertical AI?
Vertical AI refers to specialized systems engineered around the nuanced challenges of a single industry or narrow function. Unlike general purpose tools, they're trained and fine tuned on deep, domain specific data such as medical imaging records, legal case law, and proprietary manufacturing telemetry, allowing them to execute complex, context aware tasks with precision and built in regulatory alignment.
Their value is contextual accuracy and out of the box relevance: sectors like healthcare, finance, and legal services can deploy these systems to automate complex decisions and reduce compliance risk without extensive custom prompt engineering.
Which One Is Winning?
Across 4,395 U.S. and Canadian VC financings totaling $186B in 2025, vertical startups captured 53% of deal volume and 30% of capital deployed, a capital gap driven almost entirely by a handful of horizontal mega rounds like OpenAI's and Anthropic's raises. (Euclid Ventures, The Vertical Report 2026)
Harvey (legal AI) is the clearest proof point. It crossed $190 million in annual recurring revenue in January 2026, up from $100 million the previous August, and by March 2026 had raised $200 million at an $11 billion valuation. By September 2026 that figure had climbed further, to $550 million raised at a $15.5 billion valuation. (CNBC)
Sierra (customer service AI), founded by former Salesforce co CEO Bret Taylor, raised $950 million in May 2026, pushing its valuation above $15 billion. (TechCrunch)
The pattern holds up under scrutiny: capital and adoption are both moving toward systems built for one job done exceptionally well, not many jobs done adequately.
Where Quantum Fits Into This Divide
Quantum computing doesn't change the calculus for horizontal AI. Drafting, summarizing, Q&A, and code generation involve no exponential search space to prune and no molecular energy calculations to run. Classical GPU and TPU infrastructure already handles LLM inference efficiently, and bolting quantum hardware onto that pipeline adds cost and latency with no accuracy upside. The real bottleneck for horizontal AI sits elsewhere: OpenAI, Anthropic, Google, and Microsoft are pouring investment into training data quality, model architecture, and agentic reasoning, not quantum acceleration, because the constraint on LLM performance has never been raw compute. It's data and alignment.
Vertical AI is different, because vertical problems are often exactly the kind quantum computing is built for:
Quantum mechanical problems, in pharma, materials science, and chemistry, involve simulating molecular interactions that classical AI can only approximate. Algorithms like VQE (Variational Quantum Eigensolver) and QPE (Quantum Phase Estimation) calculate molecular ground state energy and eigenvalues with chemical accuracy classical methods can't match.
Combinatorial optimization problems, in finance, logistics, energy, and manufacturing, involve NP hard search spaces where classical AI returns "feasible" rather than optimal answers. QAOA (Quantum Approximate Optimization Algorithm) and quantum annealing are built precisely for this: portfolio allocation under risk constraints, vehicle routing across thousands of stops, production scheduling with hard operational limits.
This is exactly the layer Bloq Quantum is built for:
Hardware access without hardware ownership: Bloq's platform connects to cloud quantum hardware from multiple providers through a single interface, so enterprises can test which backend actually fits their problem before committing budget, rather than negotiating separate access and separate learning curves with each vendor.
A faster path to a first proof of concept: Rather than a ground-up quantum research project, Bloq's no-code environment lets a team upload its own data, select an algorithm from a pre-built library, and run it directly by cutting the setup and experimentation time that would normally take days down to minutes. That's enough to get a real, working comparison against your current classical baseline, before deciding whether the workflow is worth scaling further.