May 3, 2024 · 45m · mad
The $9B Startup Going After Snowflake and Databricks | Renen Hallak, CEO of VAST Data
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of The MAD Podcast, host Matt Turck interviews Renen Hallak, CEO and Founder of VAST Data, discussing how the $9 billion startup achieved hyper-growth and cash-flow positivity while building an all-in-one AI data platform competing with Snowflake and Databricks.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Matt holds 21.6% of the talking time here. How this is scored →
Matt as informed peer 3.1 Guest teaching 3.0 Guest disagreement 0.2 Matt pushing back 0.9
speaking balance: gold is Matt, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 3:12 Direct timeline correction
Renen directly checks Matt's assumption that the company has been around a short time, politely interrupting to clarify it has been eight years.
Hardest push from Matt ▶ 19:45 Calling out physics hyperboleMatt refuses to let Renen's sweeping statement pass unchallenged, interrupting to explicitly ask 'are you breaking the speed of light?'
Biggest teaching moment ▶ 10:34 Explaining AI storage mechanicsAfter Matt admits he is not a storage expert, Renen breaks down why historical tiered storage fails modern random-access AI workloads.
Matt holds his own ▶ 20:05 Mapping the ecosystem architectureMatt displays impressive industry expertise by correctly categorizing VAST's software integration points across hyperscalers, GPU clouds, and HPE hardware.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Matt as informed peer | Guest teaching | Guest disagreement | Matt pushing back | Why |
|---|---|---|---|---|---|---|
| Welcome and Overview of VAST Data's Mission | 2 | 1 | 0 | 0 | Matt introduces the show, highlights VAST Data's recent $9B valuation, and asks Renen for an overview of the platform and revenue metrics. Renen collaboratively outlines the platform's financial growth and cashflow status. | |
| Stealth Mode History and Positioning for Generative AI | 2 | 3 | 1 | 1 | Matt probes into VAST's stealth mode history and positioning for generative AI. Renen politely corrects Matt's timeline assumption and details how their platform was engineered for GPU-hungry unstructured data. | |
| The Strategic Advantage of Starting Late | 3 | 3 | 0 | 1 | Matt asks how VAST can compete against incumbents like Snowflake and Databricks. Renen explains that starting late allowed VAST to design specifically for deep learning without legacy tech debt. | |
| Unpacking the Disaggregated Shared Everything Architecture | 3 | 4 | 0 | 1 | Matt admits he is not a storage expert while attempting to explain traditional storage tiering. Renen educates him on how AI workloads require collapsing traditional storage pyramids. | |
| Technical Mechanics of DASE Architecture | 3 | 5 | 0 | 1 | Renen provides a deep technical breakdown of the Disaggregated Shared Everything (DASE) architecture and its expansion from DataStore to DataBase. Matt listens and asks clarifying questions as Renen explains multi-protocol database capabilities. | |
| The VAST Data Engine and Global Data Space | 3 | 4 | 1 | 3 | Renen describes the Data Engine and global Data Space, cheekily claiming VAST enables users to 'break the speed of light'. Matt immediately challenges the hyperbolic claim before Renen clarifies that compute is moved instead of massive data sets. | |
| Deployment Flexibility and Full Software Stack Control | 5 | 2 | 0 | 1 | Matt demonstrates sharp domain knowledge by mapping out VAST's software layer across cloud providers, specialized GPU clouds like CoreWeave, and HPE on-prem partnerships. Renen confirms and expands on edge compute access. | |
| Product Management Strategy and Customer-Led R&D | 4 | 3 | 1 | 1 | Matt asks a targeted product management question regarding resource allocation between legacy storage and new database offerings. Renen details VAST's lean 2-PM structure and customer-led R&D philosophy. | |
| The Grand Vision: An Operating System for AI | 3 | 4 | 0 | 1 | Matt asks what alternative stack users would need to assemble without VAST. Renen outlines how legacy tools designed decades ago fail modern AI applications and presents VAST's vision for an AI operating system. | |
| Go-To-Market Strategy and High-Density Data Clients | 4 | 3 | 0 | 1 | Matt questions how a young startup successfully pitches massive enterprise customers. Renen explains targeting high-density petabyte clients like quantitative hedge funds and life science research centers. | |
| Channel Alliances with HPE and NVIDIA | 4 | 2 | 0 | 0 | Matt brings up key channel alliances like HPE and NVIDIA. Renen elaborates on their deep technical integrations with NVIDIA DPUs, SmartNICs, and GPUs. | |
| Early Commercial Hiring and Sales Execution | 3 | 3 | 0 | 0 | Matt asks how a technical founder learned to recruit and manage high-performing sales talent. Renen shares lessons learned from his prior tenure at EMC. | |
| Reactive Schedule Management and Field Leadership | 4 | 3 | 0 | 1 | Matt asks about blending bi-continental engineering and sales cultures across Israel and the US. Renen contrasts Israeli speed and quick iteration with US long-term enterprise building. | |
| Audience Q&A: Sales Insights and Customer Listening | 1 | 2 | 0 | 0 | An audience member asks about unexpected sales insights. Renen highlights that active listening beats aggressive pitching before Matt closes the episode. |
Statements from this episode (29)
VAST Data is expanding platform to include compute functionality
“We're adding compute functionality, basically trying to be that middle of the sandwich, ah, taking advantage of latest and greatest in hardware underneath us and to give Applications, the easiest APIs on top of us.”
VAST Data has sold over $1 billion worth of software
“Revenue, we've sold more than a billion dollars worth of software so far in the last few years.”
VAST Data is growing revenue 2.5x to 3x year-over-year
“Growth, we've been growing at about two and a half to three X year over year.”
VAST Data has been cash-flow positive for nearly five years
“Efficiency, we've been cashflow positive on average for four and a half, five years now.”
XtremIO reached $3 billion in sales two years after EMC acquisition
“We hit three billion dollars in two years as part of EMC, of course.”
R&D makes up more than half of VAST Data's workforce
“So R&D is more than half the company and even within the go-to-market teams, I would Say more than half the people are sales engineers and technical people.”
Starting later than Snowflake and Databricks was VAST Data's biggest advantage
“And so I think our The biggest advantage versus those other companies that you mentioned is that we started late, and that allowed us to see this deep learning problem in front of our eyes versus machine learning, again, a lot smaller, a lot slower, and it als…”
AI model training requires fast access to all historical data
“AI doesn't conform to that model. You need fast access to everything in order to build an AI model.”
Legacy scale-out storage systems relied on shared-nothing sharding
“So every scale out system that I am aware of before VAST is based loosely speaking on a concept called shared nothing sharding.”
VAST scales capacity independently from performance via NVMe over Fabrics
“We put the drives on one side of the network, we put the logic on the other side, and we leverage a new protocol called NVMe over Fabrics to make it look like All of those drives are directly attached. That allows us to scale capacity independently from perfor…”
VAST Data achieves stateless compute by exposing data across the network
“Every node can now see the entirety of the data set. Data on low cost flash, metadata on storage class memory, all on the other side of the network, such that they become stateless, And they don't need to talk to each other anymore, and so now you have the mos…”
VAST Data enables native SQL querying directly on Parquet objects
“You can write a parquet object and then query it using SQL natively off of our platform without needing any higher layers on top.”
Row and column database complexity stems from legacy hard drive limitations
“Row-based databases and column-based databases. That all stems from storage. It stems from hard drives needing to be sequential. If we can give different views into the same information, then you don't need that complexity.”
Strict data laws drive the need for global federated AI training
“In some cases, it's not allowed to move data out of Germany, but you can run a federated training job that's global, and so that again gives us an ability to do things that couldn't be done before.”
VAST Data targets organizations managing 100 petabytes of data
“If it's a hundred terabytes, that's not interesting to us. If it's a hundred petabytes, it's really interesting.”
VAST Data builds its software stack internally instead of using open-source
“Everything is our own or almost everything is our own. We don't use open source nearly as much as other companies do.”
VAST Data employs only two product managers for 750 employees
“You'll find we're 750 people. We have two product managers in the company.”
VAST Data allocates up to 20% of R&D to future projects
“We always try to have at least 10 to 20% of R&D On advanced development, on future projects, have 10 to 20% of R&D interfacing with the field, fixing problems, and adding small features that are required, and then the bulk of the team is, of course, building t…”
VAST Data spent up to 18 months designing before writing code
“It took us about a year, year and a half before we really started writing code in earnest because we went through those design phases.”
VAST open-sources peripheral plugins while keeping its core platform proprietary
“We open source things on the periphery. If there's a plugin, if there's an interface, if there's a way to connect to other solutions, that piece we will open source. But the core of the platform is proprietary.”
Operating large AI clusters currently requires PhD-level computer science expertise
“Today, in many cases, you still need a PhD in computer science to operate a large AI cluster.”
VAST Data sold an exabyte of storage to a 30-person company
“We just sold an exabyte to a company with less than 30 people.”
HPE standardized its enterprise file storage offering on VAST Data
“HPE has tens of thousands of sellers out there, and they've decided to standardize on VAST as their file offering and soon to be other parts of their stack.”
VAST Data considers NVIDIA its strongest and most natural partner
“NVIDIA, I would say, is our best and most natural partner.”
VAST Data reached a $9 billion valuation by consistently tripling growth
“The fact that we're worth nine billion dollars now, which is staggering, is because we keep tripling.”
VAST Data's CEO ignores Slack entirely, relying on calls and texts
“I don't open Slack. I'm probably the only person in the company that doesn't use Slack, but if something is important enough, someone will text me or call me, and that's how I run my day.”
VAST's US staff works remotely while Israeli engineers work in-office
“For example, here people work from home mainly. In Israel, everybody's in the same office together.”
VAST Data explicitly operates without hierarchy or internal titles
“As you grow, and as you grow at this pace, you have to explicitly tell people that you can leave a meeting in the middle, and that you shouldn't waste time on internal crap, and that titles don't matter within the company, and that we don't have a hierarchy. W…”