Although 2022 and the first part of 2023 has been marked by notable advancements in cloud technology – mostly via AI integrations – there is still considerable restraint in today’s venture capital-backed ecosystem.
As part of our Castles in the Cloud project, which maps the activity in the venture capital-funding startup ecosystem with that of the Big 3 Cloud providers, we compile a database of fundraising at the end of each calendar year. We then identify and analyze trends that continue to play out in the immediate months following, and look for indications of how the rest of the year will go. In 2022, venture capitalists deployed $17.5B across the 410 companies we tracked.
Unsurprisingly, that’s a big drop from the banner year of 2021, which we wrote about last year. While indicating a sharp decline, this number approximates our 2020 cloud funding numbers.
That figure doesn’t translate to lower activity, however. Many venture capital firms report they are just as busy as prior to the market correction, albeit with a focus on smaller checks for early-stage companies. During the latter half of 2022 and the first few months of 2023, seed deal velocity increased at many firms, while mid-late stage companies wait to grow into existing valuations and meet higher fundraise targets around customer traction.
Of course, VC funding is just one metric to understand how the cloud ecosystem is evolving. Technological advancements, partnerships between major powerbrokers, and vertical specialization are all components of the current wave.
During this moment of lower fundraising activity, let’s look at a few fundamental trends impacting the core of the cloud industry. Along with this essay, you can also listen to my colleague Jerry Chen’s discussion about these trends on the Greymatter podcast.
- Impact of Artificial Intelligence/ML
- Vertical Specialization
Trend 1: The Rise of AI/ML in Cloud
The latest (and strongest) wave of excitement over AI/ML is playing out within the entire cloud ecosystem. The end of 2022 and 2023 saw major advancements in AI, with foundation models and their applications like Dall-e, Stable Diffusion, and ChatGPT entering mainstream public consciousness.
AI/ML continues to be a huge focus for the cloud players, across traditional tool providers (AWS Sagemaker), foundation models themselves (Google Bard), foundation model training platforms (Amazon Bedrock), and in the form of strategic partnerships/funding (Microsoft and OpenAI, Google and Anthropic).
Moreover, the AI/ML ecosystem also represents a platform shift that may resemble the cloud in many ways.
As has been discussed widely in the field, decades of AI research and development are now culminating into a new era of technology. Following previous tech transitions of mobile and cloud, AI has matured to the point where it is becoming the enabling platform tech of today. Similar to the way the Big 3 established the cloud-provider paradigm, there is great potential for AI foundation models to provide accelerant capabilities for businesses to operate and scale. Further, as the shift from on-prem to the cloud provided a window of opportunity for startups to disrupt incumbents, foundation models may also offer a similar entry point.
Like the cloud, multiple AI layers are developing:
- A concentrated group of core foundation models provided either as a service to others or used to produce a company’s own applications. This includes OpenAI, Anthropic, Cohere, and Stability AI
- A middleware layer of services designed to make the foundation model easier to use, comparable to cloud’s core capabilities like compute and storage. We’re seeing this in examples like 1. LlamaIndex for data indexing, ingestion, and query, 2. Langchain and Fixie for agent development, 3. embeddings stores like Pinecone and Chroma, and 4. monitoring tools like Honeyhive and Helicone.
- Applications built on top of foundation models like Jasper, EvenUp, Typeface, Tome, Treat, Harvey, Algolia, and many more.
Some companies like Adept.ai, Inflection, Cresta, Github CoPilot, Adobe Firefly, and Character.ai (several of which raised large rounds in 2022) span these groups, delivering applications built on top of their own in-house foundation models. They believe owning the model architecture will give them the control and customization to build a superior end application.
The development of this ecosystem has sparked significant debate around which layers will accrue the most sustained value as the underlying model provider platforms improve and expand their capabilities. Ultimately, large companies will be built in each layer. However, building in the middle layer in particular will require a strong north star, agility, and strong customer/community focus as the layers above and below evolve.
Much of the AI hype comes from consumer applications and startups leveraging the models, but (as in any sector), in time most of the spend on infrastructure will come from enterprises.
In our conversations with enterprise organizations, there is overwhelming interest in LLMs. However, they are far from production in most cases. Most enterprises are early in understanding the technology, meaning they are just beginning to define use cases for their business. After this definition stage, the next step will be focused on internal POCs and showing what could work, prioritizing model capabilities, query decomposition, customization, and speed of iteration. Then, robustness concerns like evaluation, scalability, inference latency, data privacy, and cost will rise to the forefront before models can go into production. Right now there is particular interest in shrinking down models via fine-turning, narrower task definition, and other techniques to improve cost and latency.
AI’s impact on the nature of software development, business models, and strategic partnerships seems to become more pronounced on a daily basis. We’re even starting to see competition aimed at becoming a viable fourth cloud behind the Big 3.
Oracle and NVIDIA
While still trailing the Big 3, Oracle Cloud Infrastructure (OCI) has improved their offering over the past few years, particularly in AI/ML.On March 9th Oracle announced in their earnings report that cloud infrastructure spend was up 55%, representing more than 20,000 customers and several with $1B+ contracts.
Oracle has made intelligent moves across their portfolio, combining lift and shift capabilities, strong third-party integrations, first-party connectivity with applications like Oracle ERP, and more vertical offerings (see Vertical-Specific section below). However, their most notable work might be in AI/ML and their partnership with NVIDIA.
NVIDIA’s GPU suite has provided the hardware backbone for ML, and the company recently announced their AI Foundations set of cloud services for generative AI across text, images, biology, and more. AI Foundations combines cloud services, pre-trained foundation models, frameworks, and optimized inference engines to help enterprises customize foundation models to their own data. At its core is NVIDIA DGX Cloud, which was released first on OCI (with Azure, GCP, and others in development). In fact, Adept is using OCI to develop their product.
Given the importance of AI/ML to the cloud going forward, it will be interesting to track Oracle’s and NVIDIA’s offerings over the next year.
Trend 2: Vertical Specialization
Another fundamental trend affecting the cloud landscape is the burgeoning role of the “industry cloud.”
Industry clouds bridge the gap between technology and solution, combining SaaS, PaaS, and IaaS capabilities tailored to a specific vertical into a single offering. By tightly coupling these composable cloud capabilities together, a solution can directly address vertical business challenges, effectively turning a cloud developer platform into a business platform.
Companies across the cloud ecosystem are offering industry cloud solutions. The cloud providers themselves offer individual services in areas like life sciences (Amazon Omics, Microsoft Genomics, GCP Cloud Life Sciences) and financial services (Amazon Finspace, GCP Open Banking API), but ERPs, ISVs, and SIs do much of the heavy lifting to develop on top of cloud platforms to tailor solutions to customer segments. In a 2022 Gartner survey among NA and EMEA enterprises, 39% of respondents had started the adoption of ICPs, with an additional 14% in pilots. Some startups have begun marketing around the concept over the past few years. For example, Fabric, one of this year’s new unicorns, wrote about the need for a flexible, modular, and scalable “AWS for commerce” platform role they seek to fill.
The caveat to this approach is the risk of creating vendor lock-in, so customers are still looking for options that provide a degree of modularity. In banking for example, an industry-specific solution will meet banking compliance and security requirements out of the box, exceeding typical cloud standards that would require heavy configuration.
This industry cloud trend corresponds to a renewed interest across venture in vertical software. As software and cloud infrastructure continue to grow in complexity, industries with specific needs and workforces specialized in areas outside of software need tailored solutions.
As we noted last year, this specialization is often a startup’s edge. As advancements in AI and the proliferation of industry-specific clouds are contributing to the overall growth and complexity of the entire cloud ecosystem, the need for continually-evolving security tools has also increased. Small, hyper-focused teams may be best suited to serve these needs.
Trend 3: Security
Security remained a hot category in 2022, recording the highest funding ($3.2B) among tracked companies and ranking among the highest categories by number of new unicorns (13). It’s not a surprise given Gartner ranks cloud security specifically as the fastest-growing security category, while other categories affected by the cloud such as application/developer security and data security/privacy rank as the next top two in YoY growth. While the total funding figure is highest among this year’s categories, it does represent a drop compared to last year in parallel with the rest of the market.
Startups looking for opportunities in this sector should bear in mind that while security budgets remain strong, enterprises are placing a higher emphasis on platform consolidation – as well as on security companies that replace existing budget lines, as opposed to those seeking net new spend. Correspondingly, investors are wary of potentially overcrowded segments that may not be standalone.
As far as cloud-specific security goes, companies like Wiz are growing at unprecedented rates. The secular growth of cyber overall is reflected in the spread of new unicorns across surface areas and markets – email (Abnormal), browser (Island), API (Salt), network (Perimeter81), attack surface management (JupiterOne), and MDR (Expel). Additionally, the rising sophistication of attacks and complexity of enterprise environments alongside increased regulatory scrutiny may also play a role in market growth.
Interest in supply chain, application security, and Software Bill of Materials (SBOMs) continues to grow as the results of EO 14028 take shape in the years following the SolarWinds and Log4j incidents. The prosecution of Uber’s former CSO in the wake of a 2016 data breach made it clear that in the eyes of the government, “Companies have a responsibility to protect data and to alert customers and appropriate authorities when data is stolen by hackers.”
One particular area of opportunity we expect to see in the next few years is the battle to become the security data platform, unifying data across disparate sources and enabling efficient SIEM capabilities on top. Based on our conversations with practitioners, their cloud SIEM journeys typically follow a similar path, starting with the need to pull events out of Cloudtrail logs and combine them with the appropriate context. After finding something like AWS Security Hub inadequate, they attempt to roll their own, but soon discover the challenge of stitching together big data architecture, data catalog, and data schema to search Cloudtrail and other logs in AWS. Moreover, it is nearly impossible to pull this off in a scalable way.
Given the explosion of security data, these practitioners are looking to more modern architectures like data lakes and lakehouses to provide a scalable data solution that solve the high ingress/egress cost of cloud logs while incorporating appropriate context. Snowflake, Databricks, Clickhouse, and S3 have all written about serving as the underlying data stores for decoupled SIEM detection layers on top. As I wrote previously, a semantic layer on top would make it easier for practitioners to hunt through data across multiple sources, and empower security professionals less deep in the nuances of cloud.
Looking ahead, we expect aggregate 2023 VC funding to stay at levels commensurate with current macroeconomic conditions. The financial (and to an extent, cultural) reset within Silicon Valley and in other startup hubs will persist, with the onus on many companies to catch up to their valuations. Unsurprisingly, startups that do need to raise can expect comparatively tempered round sizes at lower valuations than what was the norm in 2021 or early 2022.
Still, we are encouraged by the level of activity across many aspects of cloud and remain optimistic that the sector will continue to grow. We expect AI to be the dominant driver in the creation, growth, and funding of companies across categories. Additionally, the ongoing expansion of security needs and the rise of vertical-specific clouds and software will provide entry points for highly specialized startups.
We also believe it’s a good time to be building a company. Founders who have internalized the fallout following a period of relatively unrestrained capitalization and growth are historically more focused and deliberate. While there is always a level of risk to joining a startup, companies at the earliest stages may currently have the advantage in attracting talent during this period, as layoffs and hiring freezes at larger organizations persist.
Like last year, we believe it is possible for strong, private cloud companies with scale to go public – but not until public markets settle. We remain optimistic and are actively looking to partner with the best founders building in the cloud opportunities above.