Agentic AI Comes into Focus: What We Learned from Q1 2025 Earnings

As AI gains significant traction in the enterprise, we’re analyzing quarterly earnings calls to better understand how key players both up and down the AI stack and across different sectors are adapting their strategies, products, and pricing models to compete. 

In our 4Q24 report, companies were focused on pressure-testing AI in the enterprise. In 1Q25, we were surprised by how quickly the tune had shifted. AI has moved from an experiment to a core strategy across the tech stack, and almost every company is telling an AI story about their business.

As we did last quarter, we looked at 10 of the largest publicly-traded companies, 5 major cybersecurity and infrastructure software companies, and 4 US-based hyperscalers. Here are some of the takeaways that interested us the most.

Apps

Practically every company is embedding AI directly into its core products—and, with advances in reasoning models, they’re all racing to add autonomous agents to their roadmaps. As the market fills with AI apps, two major differentiators are emerging: exclusive access to unified customer data and the depth of in-house AI talent.

Embedding vs. standalone AI. Companies, particularly in the SMB and mid-market, are embedding AI into their core product suite as they try to get customers hooked on AI—which is a departure from the SKU-based experiments of the prior year. That said, companies generally offer more agentic products as standalone SKUs.

• "We will embed AI into every hub and every part of the platform. One product, one AI-first customer platform, no bolt-ons." —Hubspot
• “The next evolution of the Atlassian system work is officially here as we've started making a shift from standalone products to a vision of apps and agents with Rovo (AI product) at the center of everything." —Atlassian (Note that Rovo was previously a $24/month/seat add-on.)
• Microsoft rolled out M365 chat in January to provide basic Copilot access to all commercial customers, but they also maintained a standalone copilot SKU with expanded features.

Billing by usage is on the rise. As companies embed AI into their products, they’re making the switch to usage-based billing very quickly. A common strategy is to give customers a taste of AI features included in their seat-based subscription and then charge on a consumption basis as they scale up usage.

• “We want to make it super easy for our customers to start using and to scale with agents. . . . Pro and Enterprise customers will have an allotment of credit that is included within their subscription. That means when they start using customer agents, they can do that without adding additional costs. And as they increase usage and they want to scale, they can buy credits and either buy pay-as-you-go or packs of credit.” —Hubspot

Reasoning models set the stage for agents. Advances in reasoning models like OpenAI o3 and Grok4 promise to provide a step-change in productivity as they can handle complex, multi-step tasks. Agentic products typically require more intensive LLM usage and can demand higher prices—something to watch as more agents come to market. 

Unified customer data favors incumbents. Incumbents are highlighting the value of unified customer data in their platforms, and some have even started to restrict data access for third-party apps in an attempt to lock out competitors. Many AI startups are building on top of core systems of record, and these aggressive actions may force earlier-stage companies to rethink their strategy.

• Salesforce blocked third parties from indexing data from Slack, an important component of enterprise search. Their Agentforce offering crossed $100M of AOV in Q1 after entering GA in October.
• “We unify structured, unstructured, and external data at scale, giving our AI a richer, more complete foundation." —Hubspot
• "Our customers trust that Workday's AI is powered by the largest and cleanest finance and HR data set." —Workday

Demand for great AI talent is white-hot. Driven in part by the perception of a more favorable regulatory stance on M&A from the current administration, M&A is emerging as a top method for turbocharging companies’ AI capabilities and competing in a broader talent war. In recent months, ServiceNow acquired Moveworks for $2.85B and Google acquired personnel from Windsurf for $2.4B.

• "Moveworks does bring a lot of good AI expertise, the 500 people we get instead of us hiring one at a time." —ServiceNow

Infrastructure

As most infrastructure companies continue to benefit from their vendors’ accelerated cloud and data migrations, they’re also moving away from folding AI support into their existing feature set in favor of building new products to support AI in deployment.

Infrastructure modernization surges as a big priority for customers. Many enterprises are re-initiating cloud migrations and upgrading data systems in order to effectively invest in AI, which is a big tailwind for infrastructure companies built around modern technology stacks.

• "The urgency to adopt AI is omnipresent in all of our customers. It no longer seems to be a choice. . . . We're seeing customers who were previously delaying their cloud migrations are now reaccelerating their investment." —Palo Alto Networks [emphasis ours]
• Snowflake grew weekly usage of AI/ML products to over 45% of total customers, growing over 2x y/y—underscoring the move to cloud data warehouses.

Product suites expand to serve AI in deployment. The launch of new agentic capabilities has turbocharged this trend, as agents take actions without a human in the loop and create new security and monitoring problems.

• Crowdstrike launched an AI model scanner and an AI security dashboard.
• Datadog is investing both down the stack near GPUs and up the stack toward agents. Over 12% of Datadog customers now use AI integrations such as LLM monitoring.

AI agents augment the value prop of infrastructure companies. Infra companies are building AI add-ons to help analyze and activate the rich data sets created by their existing products. Similar to comments from last quarter, incumbents believe their access to a broad surface area will enable their agents to have the context and access to complete end-to-end tasks such as triaging a security threat and patching vulnerabilities.

• SentinelOne’s AI security analyst—PurpleAI—is now being attached to 25% of new customer deals.

Hyperscalers

Cloud providers continue to benefit from the AI boom, since they rent access to the GPUs that power AI models in training and inference. As demand continues to explode, they’re all vying for a larger share of the pie and trying to differentiate by highlighting the capabilities of their various model platforms and engaging in strategic partnerships with AI labs.

Demand for compute is still white-hot. Hyperscalers are still struggling to keep up despite massive investments in capacity. 

• “Amazon's AI business is a multibillion-dollar annual run rate business that's growing triple-digit percentages year-over-year,” but growth is being held back because “as fast as we actually put the capacity in, it's being consumed.” —Amazon
• “We had hoped to be in balance by the end of Q4 (June ‘25). We did see some increased demand . . . so we are going to be a little short still.” —Microsoft
• Microsoft also noted that tokens processed—a proxy for LLM usage—increased 5x y/y in the first quarter.

Emphasis on both broad model availability and success of first-party models in given categories. Hyperscalers understand that GPUs will become more available over time and trend towards commoditization, so they have looked to their offerings of different AI models as a way to differentiate. Google highlighted Gemini 2.5 and Veo (video model) alongside Llama 4 (Meta) and Claude (Anthropic). Microsoft’s relationship with OpenAI continues to be a strong point of differentiation, but over the past year, Microsoft has emphasized that the model layer will trend toward commoditization over time.

• “We brought the latest models from OpenAI along with new models from Cohere, DeepSeek, Meta, Mistral, Stability to Foundry… we’ve expanded our Phi family of SLMs with new multimodal and mini models… Phi has been downloaded 38 million times.” —Microsoft
• “Our Vertex AI platform makes over 200 foundation models available . . . including Gemini 2.5 Pro, 2.5 Flash, Imagen 3, Veo 2, Chirp and Lyria, plus open source and third-party models like Llama 4 and Anthropic.” —Google

Proprietary silicon and full-stack tooling are emerging as additional points of differentiation. Google launched its inference-optimized TPU and an open source Agent Development Kit, while Amazon is scaling Trainium2 and positioning Bedrock, an AI development suite, as a best-in-class managed platform.

• “Trainium2 offers 30% to 40% better price performance versus other GPU-based instances. It’s starting to lay in capacity in larger quantities… Amazon Bedrock is our fully managed service that offers a choice of high-performing foundation models . . . and the most compelling set of features to build high-quality generative AI apps.” —Amazon
• “Our Agent Development Kit is a new open source framework to simplify the process of building sophisticated AI agents and multi-agent systems… Ironwood, our seventh-generation TPU and most powerful to date, is the first designed specifically for inference at scale. It delivers more than 10× improvement in compute power over our recent high-performance TPU while being nearly twice as power efficient.” —Google

In short: AI is becoming the center of gravity across the tech stack. Q1 2025 earnings calls showed incumbents scrambling to embed autonomous agents, infra vendors riding a fresh cloud-modernization wave, and hyperscalers straining to meet surging compute demand while differentiating through silicon and model ecosystems. Unified customer data and scarce AI talent are emerging as the differentiators (for now!), and usage-based pricing is turning early curiosity into durable revenue. We’ll keep tracking how AI reshapes public companies, returning with fresh insights after Q2 2025.

Thanks to Justin Kahl and Ryan Allen for their contributions to the research and insights. 

More from the a16z Growth Compendium

How 100 Enterprise CIOs Are Building and Buying Gen AI in 2025

By Sarah Wang , Shangda Xu , Justin Kahl , and Tugce Erten

To complement the hard numbers from earnings with a forward look at demand, we also surveyed 100 CIOs across 15 industries about how they’re building, buying, and budgeting for gen AI. The report distills 16 key shifts—from budget reallocations and frontier model differentiation to procurement processes and the rise of AI apps—that highlight where enterprise leaders still have concerns about gen AI and where they’re doubling down. A helpful complement for understanding how the public market supply we discuss above meets up with buyer demand. 

Read here

Connect with a16z Growth

Want to stay informed on the latest from a16z Growth? Subscribe here to receive these LinkedIn newsletters regularly. 

And if you’re interested in in-depth theses, frameworks, benchmarking, and tactics for finding and growing best-in-class businesses straight from the a16z Growth investing and operating teams, check out our new Substack here.

Disclosures

You are receiving this newsletter since you opted in earlier; if you would like to opt out of future newsletters, you can unsubscribe immediately.

This newsletter is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. This newsletter may link to other websites and certain information contained herein has been obtained from third-party sources. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the enduring accuracy of the information or its appropriateness for a given situation.

References to any companies, securities, or digital assets are for illustrative purposes only and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund which should be read in their entirety.) Past performance is not indicative of future results.

Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. Content in this newsletter speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see disclosures for additional important information.