Roei Haviv

Early-stage VC @BessemerVP | Views are my own.

[Roei Haviv @roeihaviv]

https://t.co/aNMVRy8T7A

[Bessemer @BessemerVP]

Single-player AI is a dead end at work: one person, one chatbot.

@GuyWeissZZ and the @zenzap_chat team saw early that the real unlock was multiplayer, where entire teams work alongside agents, securely, with the company in control. This is why we continue to back this team, and why today's launch matters.

The agent has joined the work conversation.

Learn more from @AdamRFisher below 🧵

[Adam Fisher @AdamRFisher]

The business counterpart to Instinct/Muse is already here, and it's multi-player.

Create agents in 90 seconds, connect them hundreds of work applications, and join the thousands already using Zenzap.

After 2+ years in development, Zenzap is ready for your workplace. 🧵

[Roei Haviv @roeihaviv]

@GuyWeissZZ @zenzap_chat ⚡️⚡️

[Roei Haviv @roeihaviv]

AI agents have joined the team ⚡️

[Guy Weiss @GuyWeissZZ]

AI is everywhere, but walk into almost any company and work hasn’t changed

Meet @Zenzap_chat, the Multiplayer AI platform.

Now, people and proactive AI agents are in the same group chats,
automating workflows with shared context across business tools and teams

Everyone can bring their team and create AI agents in 90 secs (even from your phone). CISOs and CIOs love it

Thousand of teams from our early access program already transformed their day to day

AI agents have joined the team

[Roei Haviv @roeihaviv]

Agree. The worst thing is doing the work and building conviction, but buying in modestly instead of going all in, and then watching it explode. I’m okay missing it in the first place ($NVDA for me) or not holding long enough ($DDOG for me), but not really enjoying the upside when you’re already in - that’s brutal.

[Roei Haviv @roeihaviv]

@Funmentalist Amazing anaylsis @Funmentalist

[Roei Haviv @roeihaviv]

SpaceX acquiring Cursor was one of the most important milestones of the AI era. It proved that venture-scale returns can be generated at mind-blowing multiples incredibly fast.

The main question is whether this is indicative of what’s coming. And if so, how long it can last.

[Roei Haviv @roeihaviv]

A public company is priced on its execution against expectations.
A private company is priced on its vision and perceived ability to execute against it.

Those are two different measures of value. There’s no sense in comparing them.

[Roei Haviv @roeihaviv]

The most important challenge in AI is ensuring that it operates in line with human values. But the US and China have very different values.

[Roei Haviv @roeihaviv]

@iancassel @iancassel brilliant. Do you think the same applies to VC investments?

[Roei Haviv @roeihaviv]

The addressable market for tech companies used to be the software budget within other companies. Now, with both LLMs and physical AI, the addressable market is essentially all of OpEx and all of CapEx.

[Roei Haviv @roeihaviv]

Indeed fascinating. But it all rests on a few assumptions:
1) The fact that we're still at the beginning doesn't mean the current dynamic will hold. Part of the cost implosion is a land grab. Once frontier labs shift from capturing territory to defending margins the decline in prices may slow.
2) Cheaper tokens may reduce the cost of accessing intelligence, but they don't by themselves make intelligence cheaper. Tokens measure what a task consumed, not the value it created.
3) Putting inference on the edge (i.e., fridge/watch) requires a different architecture/hardware that must keep pace with the cycles of models on one end and chips on the other, which will probably drive different economics.

[Roei Haviv @roeihaviv]

@SemiconductorsX No better indicator of overdemand if this plays out.

[Roei Haviv @roeihaviv]

@jukan05 Nvidia remaining this massive while custom silicon scales alongside it shows just how enormous the AI compute market is becoming.

[Roei Haviv @roeihaviv]

Like this framing. I'd pair it with hyperscaler capex. Right now the two charts move together, which makes the thesis even stronger.
More specifically, neoclouds are probably the AI buildout's highest-beta exposure to this dynamic - fastest to reprice up in scarcity, yet fastest down when supply starts to catch up.

[Roei Haviv @roeihaviv]

@HyperTechInvest The current memory dynamic almost demands an architectural breakthrough that reduces HBM dependency and makes compute less memory-bound.

[Roei Haviv @roeihaviv]

זה נכון שהשוק הפרטי זז מהר יותר, אבל כשחברה מגייסת סכומים כאלה בסוף היא צריכה לייצר נזילות.

השוק הציבורי מתמחר ביצועים וצמיחה רבעונית, בעוד שהשוק הפרטי יכול לתמחר חזון ארוך טווח. זו אחת הסיבות ש-SpaceX נשארה פרטית כל כך הרבה זמן.

כל עוד Databricks יכולה למכור את החזון שלה סביב הנרטיב של AI, אין לה סיבה להפוך לחברה ציבורית, אבל בסופו של דבר לא תהיה להנהלה ברירה אלא להנפיק.

[Roei Haviv @roeihaviv]

@Speculator_io The speed is unprecedented but the scale is not (at least not yet). Railroad capex peaked ~6% of US GDP in the Gilded Age. AI capex is running ~2% today, and even the bullish projections for 2030 land at ~5.6%.

[Roei Haviv @roeihaviv]

@bits__to__atoms @SemiconductorsX This market is too big for a winner takes all dynamic. I think it will look more like the hyperscalers market.