SYSTEM OPTIMAL
LATENCY: 12msALERT: HBM3E SUPPLY CONSTRAINED
08:08:33 UTC
BC

Betsy Chernoff

Principal AI Product Marketing Manager · WEKA

2 appearances2 events0 quotes cited0 newsletters

Position Evolution

3 tracked across this operator's appearances

Same operator, on the record, on the same topic, at different points in time. Each delta below is anchored to verbatim transcript spans verified against source — no paraphrases. This is the alumni-graph moat: SemiAnalysis cannot reproduce this query because they don't have the speaker-stable corpus.

GPU cost reduction via token throughput

Hardenedconfidence 88%

At GTC Chernoff cited a 6x token-per-second gain as the headline metric. At SC25 she translated throughput gains into explicit dollar and GPU-count savings, making the business case far more concrete and accessible to buyers.

Earliest · Apr 2026 · nvidia

"what we've seen for them is that they're getting upwards of 6x improvement in tokens per second, and that's huge, right?"

Source on theCUBE ↗
Latest · May 2026 · acm-ieee

"4. 2x reduces that cost to somewhere around $2. 28 million. So $ 720, 000, or 24 GPUs, instead of a hundred. And that's the context that we want to get to."

Source on theCUBE ↗

KV cache persistent storage tier

Hardenedconfidence 85%

In April Chernoff framed the persistent KV cache tier as a new opportunity just becoming possible. By May she is quantifying it concretely in terabytes-to-petabytes terms and positioning WEKA as the definitive solution, showing growing specificity and confidence.

Earliest · Apr 2026 · nvidia

"There was nothing in between, right? So there was no having it ready for reuse and sending it back to hot state quickly. None of that existed before. So today, we now have that opportunity to fundamentally build in a meaningful way."

Source on theCUBE ↗
Latest · May 2026 · acm-ieee

"We are that persistent layer of memory, if you will. And while those two may be ephemeral, or not be persistent, we provide that persistence in a way that's incredibly meaningful. So those systems typically have about terabytes, and we can move all the way up to petabytes."

Source on theCUBE ↗

Cloud availability of augmented memory grid

Shiftedconfidence 82%

In April the focus was a single production-grade POC with Firmus on NVIDIA infrastructure. By May the emphasis moved to broad commercial availability on OCI, signaling a shift from proof-of-concept validation toward multi-cloud go-to-market expansion.

Earliest · Apr 2026 · nvidia

"we announced not only are we partnering with NVIDIA around STX, but Firmus has done a POC with us, production grade for augmented memory grid."

Source on theCUBE ↗
Latest · May 2026 · acm-ieee

"we're also announcing this commercial availability on OCI, Oracle Cloud Infrastructure."

Source on theCUBE ↗

All theCUBE appearances (2)

  • SC25 | Callan Fox & Betsy Chernoff, WEKA

    GUEST · WEKA · Principal AI Product Marketing Manager

  • NVIDIA GTC 2026 | Betsy Chernoff, WEKA, & Ace Stryker, Solidigm

    GUEST · WEKA · Principal AI Product Marketing Manager