Get Plus

Sign in to baba News

One account across the web, iPhone and Android — your subscription follows it.

or use an email code

Search stories

Type at least two characters. Results come from every newsroom baba reads.

↑↓ to move · ↵ to open · esc to close

Sign in to baba News

Sign in to keep asking. News Plus removes the daily limit.

or use an email code

CalcalistTech

Palo Alto Networks' AI Infrastructure Deal Signals Missed Israeli Opportunity

Translated & summarized from Calcalist by baba

BusinessCharged tone

Hebrew · 2 newsrooms covering

Palo Alto Networks acquired Applied Compute for $4.5 billion to enter the AI infrastructure market, focusing on customizing AI models rather than traditional cybersecurity. This strategic shift reflects a broader industry trend towards independent AI control. Experts believe Israel missed an opportunity in this burgeoning field but still has potential. The deal signifies a new phase in the AI race, emphasizing efficiency and customization.

The story in 5 lines · by baba

  • Palo Alto Networks acquired Applied Compute for $4.5 billion, entering the AI infrastructure market.
  • The deal signifies a trend of companies developing independent AI capabilities instead of relying on external providers.
  • Experts see this as a missed opportunity for Israel's tech industry in the AI model refinement sector.
  • Adapting AI models requires specialized infrastructure and expertise, fostering a new service industry.
  • The AI race is shifting from developing large models to making them efficient and customized for enterprises.

Palo Alto Networks has acquired Applied Compute for $4.5 billion, a move that surprised the market not just for its high price but for its strategic shift. Instead of acquiring another cybersecurity firm, the tech giant is entering the AI infrastructure space, focusing on technologies that enable organizations to develop and customize AI models for their specific needs. This acquisition signifies a broader trend among major tech companies to move beyond using models from providers like OpenAI and Anthropic, aiming instead for independent control over the AI powering their products and services. For Palo Alto Networks, this means developing AI models tailored for cybersecurity, enhancing their performance, and reducing reliance on external AI vendors.

Yonatan Mandelbaum, a partner at venture capital firm TLV Partners, explained that the acquisition reflects the growth of an industry focused on refining existing AI models and adapting them for enterprise use. He highlighted that the key difference lies in accessing the "weights" of a model, the parameters determined during training that dictate its behavior. "When using closed models like those from OpenAI or Anthropic, you don't have access to their weights, so you can't make independent adjustments. Open models, on the other hand, allow companies to take an existing model, improve it, and adapt it to their needs," Mandelbaum stated. He added that companies seeking control over their organizational intelligence would want the best balance of cost and performance, leading to the development of this specialized industry.

Adapting open AI models requires complex technological infrastructure, including data collection, preparation, model refinement, and operational optimization. This process demands research and engineering expertise that most organizations lack, fostering an industry of service providers. Mandelbaum sees Palo Alto Networks' acquisition of Applied Compute as a prime indicator that the post-initial-training model refinement sector is becoming a significant market. He believes Israel had a substantial opportunity in this area due to its talent pool but is "a bit late to the party," citing the acquisition by Palo Alto Networks rather than an Israeli team as evidence of a missed opportunity. However, he remains optimistic, noting that "it's not too late" and several exceptional Israeli teams are working on these challenges.

Uri Goshen, co-founder and CEO of AI21, which develops AI models for organizations, views the deal as a manifestation of a profound shift in how companies perceive AI usage, from reliance on vendors to owning their models. He explained that organizations using third-party AI services effectively pay twice: once for usage and again for the missed opportunity to improve the AI over time. In an era of AI infrastructure consolidation, control over models and their continuous improvement becomes a crucial competitive advantage. Goshen noted that companies are realizing that true advantage comes from accumulating organizational knowledge, tailoring models to company operations, and continuously enhancing them, thereby building a long-term technological asset instead of repeatedly paying external providers.

Mandelbaum raised two key open questions in this market: whether model refinement and model execution will merge into a single domain, and how companies in this space will convince clients to engage with their technology. He pointed out that Applied Compute derived significant revenue from model execution services it helped train, while many AI infrastructure providers are now offering model adaptation and refinement services. The crucial question for clients, he noted, is not the underlying technique but achieving the best performance at the lowest cost. Ultimately, the acquisition signals a new phase in the AI race, shifting from developing the largest models to making them efficient, affordable, and tailored to specific organizational needs. For companies like Palo Alto Networks, mastering this technology could be as vital as their core cybersecurity offerings. For Israel's tech industry, it represents a significant opportunity but also a reminder of the rapidly evolving market and the potential for falling behind.

CalcalistOther · Tel Aviv

Sign in to baba News

Sign in to follow newsrooms, topics and people.

or use an email code

TechTopic

Sign in to baba News

Sign in to follow newsrooms, topics and people.

or use an email code

Mentioned

Full coverage · 2 outlets
First: Globes · 7h ago

The same event, reported separately by each outlet. Open a few to compare what different newsrooms emphasize — and what they leave out.

Open the Wire