← HomeAll journeys    About    Share your journey
← United StatesThe Exclusive FeatureShare your journey
Startup Talk
The people who built United States's companies · November 2026

AI data infrastructure, flash storage software

VAST Data

Hallak's thesis that flash was too small and costly while disk was too slow, and that QLC flash plus new memory and NVMe over fabrics could end tiered storage.

Renen Hallak walked out of XtremIO in 2015 without a product. What he had was a stack of conversations. He spent months asking storage buyers and storage companies what they were stuck with, and the same complaint kept coming back: flash was too small and too expensive to hold everything, disk was too slow, so everyone lived with tiers, a fast pile and a slow pile, and people whose job was moving data between them.

Hallak knew that world from the inside. He had come to storage from computational complexity research, a field closer to mathematics than hardware, and joined XtremIO at its start in 2009 as its first engineer. EMC bought the company in 2012, reportedly for about $430 million, and Hallak rose to VP of R&D, at one point managing a team of more than 300 engineers. Then he left the safe seat in the middle of that machine.

Headquarters went to New York, near the banks, studios and labs that were the likely buyers.

His bet was specific. Cheaper QLC flash, better data reduction, a new kind of persistent memory for writes and metadata, and NVMe over fabrics could together let one flash system do the job of both tiers. In January 2016 he founded VAST Data with Shachar Fienblit, who had been CTO at Kaminario, and Jeff Denworth, who came from CTERA. Research and development sat in Israel. Headquarters went to New York, near the banks, studios and labs that were the likely buyers.

Then came the quiet part. VAST spent years building something that, if it failed, would fail completely, because a storage architecture cannot be shipped half finished. A $40 million Series A arrived in 2017, and a second $40 million round, led by TPG Growth with Norwest, Dell Technologies Capital, 83 North and Goldman Sachs, followed. When the company surfaced from stealth it had hired the veteran analyst Howard Marks as evangelist, which the trade press noted was his first full-time job since 1987. A Forbes headline put it at $1.2 billion in 2020.

What nobody could plan was ChatGPT. After its launch in late 2022, every big AI data centre needed to feed thousands of GPUs at once, and a design with no tiers suddenly looked like foresight. VAST made the Financial Times fastest-growing list in 2023, raised $118 million at about $9 billion that year, and signed a $1.17 billion deal with CoreWeave in November 2025.

In April 2026 it raised about $1 billion at a $30 billion valuation, led by Drive Capital, with an IPO a recurring rumour. The sources do not say where Hallak was born or studied, and he rarely tells personal stories in public. What they do show is a habit: he talked to customers before writing a line of company code, and then he waited years for the world to need what he had built.

The Chronology year by year

  1. 2009
    • Hallak joins XtremIO as first engineer
  2. 2012
    • EMC buys XtremIO (about $430M per one source)
  3. 2015
    • Hallak leaves
  4. 2016
    • VAST founded
  5. 2017
    • $40M Series A
  6. 2019
    • out of stealth
  7. 2020
    • Forbes headline cites $1.2B valuation
  8. 2025
    • $1.17B CoreWeave agreement
  9. 2026
    • $30B valuation

The Money Trail how it was funded

  1. 2017 Series A $40M
  2. later Series B $40M (TPG Growth lead
  3. Norwest, Dell Technologies Capital, 83 North, Goldman Sachs)
  4. late 2023 Series E $118M at about $9B valuation
  5. Apr 2026 Series F about $1B (Drive Capital lead
  6. primary and secondary) at $30B

Principal backers: TPG Growth, Norwest Venture Partners, Dell Technologies Capital, 83 North, Goldman Sachs, Drive Capital; described as Nvidia-backed by Bloomberg

The Close Calls what nearly broke it

  • 2016-19: built a new storage architecture in stealth before a product shipped
  • 2015-16: bet on QLC flash and new memory when disk-based tiering was the norm
  • 2022-23: shift from storage vendor to AI infrastructure supplier [unverified detail]

In Their Own Words from the record

Spotted something wrong? Send a correction.

Startup Talk · Preview edition · Figures are from public sources and may be out of date