(Q3 2023 Letter)
The launch of Open AI’s Generative Pre-Trained Transformer (GPT) has set in motion an arms race for artificial intelligence applications. Like any model, the competitive differentiation that an artificial intelligence model provides is rooted in the data inputted into the model. To drive meaningful value, large sets of data need to be collected and stored in a way that is then readily available for the model’s GPU hardware to access for training and inferencing. However, access to this data is constrained by technical debt in enterprise data architectures that were previously designed for the lowest cost method of bulk data storage. Because historically there was no usage requirement to access such data, bottlenecks caused by slow input and output speeds were never an issue. With the advent of artificial intelligence, the need to access data has reset the requirements for data infrastructure. Data previously relegated to “cold” bulk storage now needs to be “warm” (accessible) so that it can be processed on demand. With data generation and data collection growing at exponential rates amidst a multiyear technological catalyst driven by artificial intelligence, we see an inflection in the value proposition for businesses to optimize data management as a high technology value driver.
Pure Storage was founded in 2009 to drive a purpose-built, high technology approach to data management. While others approached data management as a commodity, Pure’s high technology approach has created a data management platform whose market share gains are poised to accelerate. In Pure, we see several paths to value creation.
First, in breaking the upfront price barrier with hard disk in March, Pure extended its runway for long term revenue growth. Due to historical upfront cost advantages, hard disk is the predominant bulk storage architecture. Because of technological advancements, constraints posed by the laws of physics, and demands from artificial intelligence applications, hard disk is no longer the lowest cost option and bulk storage is no longer a commodity market. The issue facing hard disk drives lies in their speed, reliability, and cost of ownership. Hard disk drives use spinning disk that are magnetically copied when read and written in sequential order when data is stored. Solid state drives were designed to fit in hard disk architectures and are thus subject to the control rules of hard disk protocols. Flash memory used in solid state drives is then subject to the constraints of both solid-state drive manufacturers and the hard disk drive architecture itself. The end result is that hard disks are slow and unreliable and solid-state drives require memory and compute to accommodate hard disk protocols which adds expense and creates instability. In addition to inferior performance and additional costs, both hard disk and solid-state drives require full time employees to manage the inherent system constraints. Pure, on the other hand, was purpose built to solve all these problems. The result is the only software and hardware platform that works directly on the flash semiconductor to provide greater speed, higher reliability, less complexity, and lower operating costs. Pure has long offered superior data management services, but with data management viewed as a cost center, inferior embedded solutions were slow to cede market share. As bulk storage replacement cycles come due, we see a multiyear catalyst for accelerated market share gains backed by tailwinds from a secularly growing market.
Second, Pure’s growing revenue streams represent sustainable growth that creates value. Unlike businesses selling hardware in episodic replacement cycles, Pure has a repeatable business model with a recurring storage as a service revenue stream. Subscribers receive the most modernized data management architectures without the burden of obsolescence or the need to manage hardware migrations. In a growing base of 11,550 global customers, we see meaningful net expansion opportunities. Additionally, we see more data centers moving to all flash architectures, a future expansion of co-developed products and partnerships with Nvidia, Snowflake, Microsoft and Meta, and a pipeline of new opportunities with hyperscalers and large telecom operators. We also see a shift in mix favoring subscription revenue, which carries a greater margin profile, and we believe the resulting operating leverage and earnings growth profile is misunderstood and undervalued.
Third, Pure’s portfolio sale is creating a standard in high technology data management. After first validating the market for high performance data management, Pure has since enabled the technology to fulfill larger segments of customer needs from the highest performance all the way down to the lowest price. Because every aspect of the software and hardware platform was purpose built, an automated operating system and consistent user experience create an environment that “just works”. When the alternative is dozens of architectures that require full time employees to manage, it is easy to see why new customers expand the product offering to standardize Pure across their architecture. We see Pure creating an emerging standard for data management infrastructure across a growing base of global customers.
In Pure, we see a long runway of profitable revenue growth, an emerging standard in data management, and competitive advantages and economies of scale to lower costs and grow earnings.
Comments