Most enterprise migration plans do not fail on the architecture diagram. They fail on the details nobody scoped.

Moving data off SAP, DB2, or SQL Server into Snowflake, Databricks, or ClickHouse looks clean on paper. Then a pipeline breaks mid-run and nobody knows which records made it across. Or the source system cannot take a maintenance window because it is running the business. Or an audit asks who moved what, when, and from where, and the answer lives in someone's cron job.

That gap is worth understanding, because it is where most AI initiatives quietly stall. Not on model quality. On whether the underlying data is current and trustworthy enough to act on.

On August 6 at 10 am ET, CData is running a live session on exactly this problem. Working pipelines across the systems enterprise teams actually run.

From Legacy to Lakehouse Without Breaking Production

What gets demonstrated:

Change data capture instead of full table scans. CDC reads the change log rather than re-reading the table, so the source system carries far less load and downstream targets stay current without batch windows or manual refreshes.

Hybrid and on-prem connectivity without custom scripts. Legacy systems connected to modern ones without re-architecting what is already in production.

Governed data feeding AI workloads. Models, agents, and analytics running on fresh data rather than stale copies of uncertain lineage.

Fault tolerance shown in practice. Incremental checkpointing, auto-retry, and automatic recovery, so a failed run resumes instead of restarting from zero.

If your team is standing up AI on top of data you do not fully trust, this is the layer worth watching closely.

Prefer to test it yourself first, CData Sync has a 30-day free trial.

See you there!!!!

I had a blast chatting with Jithendra Vepa, CTO and Co-Founder of Observe.AI, on The Ravit Show at MongoDB.local Bangalore.

Jithendra has a PhD in speech technology and has spent years deep in speech recognition, NLP, and voice AI. He is not someone who got into AI because it became trendy. He has been building domain-specific AI systems long before the current wave, and it shows in how he thinks about the problem.

Here is what we got into.
-- We started with Observe.AI itself. What they are building, what problem they set out to solve, and why it matters for enterprises dealing with customer conversations at scale. Observe.AI is a contact center AI platform that helps businesses analyze customer interactions, coach agents in real time, and improve performance across support and sales. More than 300 organizations use it. They process millions of support touchpoints daily.

-- We got into the wins and patterns that have stood out as Observe.AI has scaled. Customer outcomes, operational improvements, how teams are actually using the product once it is embedded. The patterns here tell you a lot about where enterprise AI is actually delivering value versus where it is still a slide deck.

-- Jithendra gave the keynote at the event. I asked him what the biggest takeaway he wanted the room to leave with was. His answer came from someone who has built a 40 billion parameter contact center LLM and trains domain-specific models instead of relying on generic ones. That distinction matters more than most people realize.

-- We closed on the signal versus hype question. His advice for founders and enterprise teams trying to decide where to place their bets right now was grounded in years of shipping, not months of experimenting.

A few things stayed with me.

Generic AI is not enough for enterprise. Domain-specific models built on domain-specific data is where the real moat lives.

The companies winning in AI are not the ones with the most models. They are the ones with the most structured access to the right data at the right moment.

Contact centers are one of the first places where AI is delivering measurable ROI at scale. What is happening there is a preview of what is coming for the rest of the enterprise.

I visited INTEL HEADQUARTERS!!!!

I was at Intel at their HQ in Santa Clara. I just cannot stop thinking about what this company pulled off. Not a sponsored post. Check the numbers below!!!!

Two years ago Intel was trading at $19.98. People were writing obituaries. This year it crossed $133 and finally broke its August 2000 peak. A record that stood for 26 years.

I want to talk about how that happened. Because the story is wilder than the stock chart.

For decades, Intel WAS the chip industry. Intel Inside was on every PC on the planet. Then came the misses. Mobile went to ARM. Manufacturing leadership went to TSMC. Apple walked away and built its own silicon. When AI hit, the money went to GPUs, and Intel, a CPU company at its core, was left out of the biggest tech cycle of our lifetime.

That was the setup when Lip-Bu Tan took over. What he did next is a case study.

He bet the company on manufacturing. The 18A process is hitting yield targets months ahead of schedule. Intel bought back full ownership of its Ireland fab from Apollo for $14.2 billion. Painful, expensive, unglamorous work.

Then 2026 happened.

- Q1 revenue came in at $13.6B against a $12.4B estimate. EPS of $0.29 when the Street expected $0.01. The stock jumped 24 percent in one day, its best session since 1987.

- Q2 was even better. Revenue up 25 percent to $16.1B, the fastest growth in over 15 years. The data center and AI unit grew 59 percent.

And the partnerships tell you where this is going. NVIDIA put in $5 billion for an equity stake. Intel's Xeon 6 is the host CPU inside NVIDIA's DGX Rubin systems. Tesla and SpaceX signed on for the Terafab AI data center project.

Here is the insight I keep coming back to. Everyone assumed AI was only a GPU story. But every GPU cluster needs a control plane, and that orchestration layer is the CPU. Intel was never out of the AI stack. The market just stopped looking.

There is also a supply chain angle nobody should ignore. Intel is the only American company that designs AND manufactures leading edge chips at scale. In a world that wants chips made closer to home, that position is unique. If the foundry bet works, this starts to look like a second TSMC.

Now the honest part. The stock is down about 30 percent from its July high. It trades near 88x forward earnings and free cash flow is still negative. The market has priced in a lot of the recovery already. The easy part of this trade is over. The execution part is just beginning.

But after 750 plus interviews with leaders in this space, one pattern holds. The companies that win are not the ones that never fall. They are the ones that rebuild from first principles after falling.

Thank you my friend Annie Shea Weckesser, CMO, Intel for the conversation, you are the best!!!!

Is Intel back for good, or is the market ahead of the story? Tell me below.

🔍 Stay Ahead in AI & Data! Join 137K+ Data & AI professionals who stay updated with the latest trends, insights, and innovations.

📢 Want to sponsor or support this newsletter? Reach out and let's collaborate! 🚀

Best,
Ravit Jain
Founder & Host of The Ravit Show