According to a recent LinkedIn post from Speedata, the company is drawing attention to a shift in AI infrastructure bottlenecks from GPU-based model inference to data processing, particularly under heavy SQL workloads generated by AI agents. The post references Databricks data suggesting AI agents now account for roughly 80% of modern database queries and argues that CPUs are not optimized for this scale of concurrent, bursty SQL activity.
Claim 30% Off TipRanks
- Unlock hedge fund-level data and powerful investing tools for smarter, sharper decisions
- Discover top-performing stock ideas and upgrade to a portfolio of market leaders with Smart Investor Picks
The LinkedIn post highlights a view that large language models increasingly rely on SQL as a “truth layer,” and that vector databases alone may be insufficient for many enterprise use cases. It further notes that machine learning and reinforcement learning pipelines depend on high-throughput SQL at massive scale, implying that data infrastructure, rather than models, may become the primary performance and cost constraint.
According to the post, Speedata sees the future AI stack as a heterogeneous mix of GPUs for AI computation, CPUs for orchestration, and dedicated analytics processing units, or APUs, for SQL and analytics workloads. This framing positions APUs as a potential solution to emerging bottlenecks in data-intensive AI applications, suggesting a growing market opportunity for specialized silicon targeting database and analytics acceleration.
For investors, the post suggests that Speedata is aligning its strategy with the perceived need for purpose-built hardware to support AI-driven SQL workloads. If these trends materialize at scale, companies offering accelerators that reduce query latency and infrastructure costs could gain traction with enterprises and cloud providers, potentially improving Speedata’s long-term revenue prospects and competitive stance in the AI infrastructure ecosystem.