Home
ログイン
会員登録
リスニング練習
リスニング練習
/
Video
/
Business Insider
/
How To Manage Token Costs While Keeping Agentic Goals On Track
How To Manage Token Costs While Keeping Agentic Goals On Track
学習モードを選択:
字幕を表示
単語を選ぶ
単語を書き直す
Highlight:
3000 Oxford Words
4000 IELTS Words
5000 Oxford Words
3000 Common Words
1000 TOEIC Words
5000 TOEFL Words
字幕 (111)
0:00
Tokenomics is not an engineering
0:01
conversation anymore. It's a P&L
0:04
conversation.
0:10
My name's Bharat Bhatia, solution
0:12
architect at the Customer Solution
0:14
Center here at Dell Technologies, where
0:16
I get to showcase Dell Technologies
0:18
newest innovations, [music]
0:19
the business outcomes they drive, and
0:22
the strategic value that they can
0:23
create. Tokenomics comes from token
0:26
economics.
0:27
So, imagine tokens are just words that
0:30
AI uses to communicate. What happens is
0:33
[music] a token, which is the basic unit
0:36
that is used for AI to run workload,
0:39
it's used to read [music] and write
0:40
text. So, the more you use, the more
0:43
cost it would become. [music]
0:45
It becomes an economic conversation. So,
0:47
hence the word tokenomics. Tokenomics is
0:50
not an engineering conversation anymore.
0:52
It's a P&L conversation. It's [music] a
0:54
mindset shift because if you think about
0:57
organizations, they've been using
0:59
chatbots. Old-school chatbots [music]
1:01
are like vending machines. You press a
1:03
button, and out comes a snack. Agentic
1:05
AI, on the other hand, is a little more
1:07
complicated.
1:08
>> [music]
1:08
>> They're like your home office agents.
1:11
They can think, they plan, they [music]
1:13
use tools, they operate autonomously,
1:16
but they run for extended periods of
1:17
time. Chatbots only consume hundreds of
1:20
tokens, whereas an agentic AI workflow
1:23
can consume hundreds of thousands of
1:25
[music] tokens, or even millions. Now,
1:28
users are thinking, "What can we do
1:30
next?" And that's why agentic AI has
1:32
come about. From an agentic AI
1:34
perspective, you are now having
1:37
autonomous agents not go do simple
1:40
tasks, but do more complicated tasks.
1:42
Those complicated tasks
1:44
consume more tokens. [music]
1:46
If they consume more tokens, then your
1:48
bill is going to get higher.
1:49
This shift means that the entire
1:52
supporting ecosystem, the hardware, the
1:54
software, the security, [music] now has
1:56
to grow up with this new agentic AI
1:58
world. This is no longer just an IT
2:01
conversation. It's a whole company
2:02
conversation. This requires new
2:04
architecture for the work [music]
2:05
itself. You need trusted data
2:07
governance, infrastructure that's closer
2:10
to make real-time decisions. Where Dell
2:12
and NVIDIA come into the picture is
2:15
through Dell's [music] desk-side agentic
2:16
AI solutions, where you can build and
2:19
run secure autonomous agents right at
2:21
your desk. Dell's desk-side agentic AI
2:24
is powered by Dell's high-performance
2:26
workstations and NVIDIA's Nemo Cloud.
2:28
Those workstations could be Dell Pro
2:30
Precision Towers or the Dell Pro Max
2:33
with a GB10 or even the Dell Pro Max
2:36
with a GB300.
2:38
This allows organizations to build and
2:40
deploy agents and govern them with
2:41
privacy controls. Think of Dell
2:43
desk-side agentic AI as a clear path
2:46
from let's try this to this is running
2:49
in production.
2:50
It converts variable cloud token costs
2:54
into a controlled infrastructure
2:56
investment.
2:57
The Dell AI Factory with NVIDIA now
2:59
extends to the edge and desk-side,
3:01
[music] so you can start small, validate
3:03
your use cases, and then grow into the
3:06
infrastructure that we provide.
3:07
>> [music]
3:08
>> It lets you run AI agents right at your
3:11
side on hardware you own, not cloud
3:13
servers you're [music] renting. It's the
3:14
same tools, the same playbook, but a
3:17
bigger stage. Here's what leaders who
3:19
are exploring agentic AI need to keep in
3:21
mind. [music] Autonomy is the promise of
3:23
agentic AI,
3:24
but use cloud as a specialist [music]
3:27
when you need to,
3:28
but run agentic AI on a local system as
3:32
your everyday workhorse. Leaders should
3:33
be more focused, not necessarily on the
3:36
cost per token. They should be more
3:38
focused on cost per [music] outcome.
3:40
Where you run AI is just as important as
3:43
which AI you run. The Dell AI Factory
3:45
with NVIDIA has [music] the pieces for
3:47
you to be successful in your
3:49
organization from the edge, the
3:52
desk-side, [music]
3:53
the data center, all the way to the
3:54
cloud. Leaders have to remember, start
3:56
local, govern early, and scale smart.
4:06
>> [music]
